CN112711713B - Interest point recommendation and display method and device, computer equipment and storage medium - Google Patents
Interest point recommendation and display method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN112711713B CN112711713B CN202110054033.4A CN202110054033A CN112711713B CN 112711713 B CN112711713 B CN 112711713B CN 202110054033 A CN202110054033 A CN 202110054033A CN 112711713 B CN112711713 B CN 112711713B
- Authority
- CN
- China
- Prior art keywords
- recommendation
- interest
- recommended
- interest point
- candidate
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 62
- 239000011159 matrix material Substances 0.000 claims abstract description 430
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 177
- 238000012549 training Methods 0.000 claims description 77
- 230000001960 triggered effect Effects 0.000 claims description 37
- 238000004364 calculation method Methods 0.000 claims description 30
- 238000004422 calculation algorithm Methods 0.000 claims description 23
- 238000012216 screening Methods 0.000 claims description 22
- 238000004590 computer program Methods 0.000 claims description 21
- 238000012790 confirmation Methods 0.000 claims description 15
- 238000013528 artificial neural network Methods 0.000 claims description 10
- 238000012163 sequencing technique Methods 0.000 claims description 9
- 238000001914 filtration Methods 0.000 claims description 8
- 238000010801 machine learning Methods 0.000 claims description 7
- 238000009795 derivation Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 19
- 238000004891 communication Methods 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 238000005457 optimization Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 4
- 238000011156 evaluation Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 235000019633 pungent taste Nutrition 0.000 description 2
- 239000000243 solution Substances 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 235000013361 beverage Nutrition 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010790 dilution Methods 0.000 description 1
- 239000012895 dilution Substances 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application relates to a point of interest recommendation method, a point of interest recommendation device, computer equipment and a storage medium. The method comprises the following steps: determining an initial range based on each initial position corresponding to each object to be recommended; obtaining a candidate range of interest points to be recommended based on the initial range; performing matrix building based on historical recommendation degrees of the interest points to be recommended and the historical interest points corresponding to the objects to be recommended to obtain an original interest point recommendation degree matrix; performing matrix decomposition prediction based on the original interest point recommendation degree matrix to obtain a target interest point recommendation degree matrix; determining candidate interest points from the interest points to be recommended based on the interest point attribute information corresponding to the objects to be recommended, and searching candidate recommendation degrees corresponding to the candidate interest points from a target interest point recommendation degree matrix; and determining target recommendation interest points based on the candidate recommendation degrees, and sending the target recommendation interest points to the terminals corresponding to the objects to be recommended. By the method, the accuracy of recommending the interest points of multiple persons in the map can be improved.
Description
Technical Field
The present application relates to the field of computer technologies, and in particular, to a method and an apparatus for recommending and displaying a point of interest, a computer device, and a storage medium.
Background
With the development of internet technology, a point of interest (POI) recommendation technology appears, and a POI is a term in a geographic information system, which generally refers to all geographic objects that can be abstracted as points, especially some geographic entities closely related to people's lives, such as schools, banks, restaurants, gas stations, hospitals, supermarkets, and the like. The main purpose of the interest points is to describe the addresses of the things or events, so that the description capability and the query capability of the positions of the things or events can be greatly enhanced, and the accuracy and the speed of geographic positioning are improved. At present, the recommendation of the interest points is performed by sequencing single conditions such as distance, price, score and the like to recommend the users, however, when the interest points gathered by a plurality of users need to be recommended, the accuracy of the recommended interest points is low due to the method of recommending according to the single conditions, and the requirements of the plurality of users cannot be met.
Disclosure of Invention
In view of the above, there is a need to provide a point of interest recommendation and presentation method, apparatus, computer device and storage medium capable of improving accuracy.
A point of interest recommendation method, the method comprising:
acquiring initial positions and interest point attribute information corresponding to objects to be recommended, screening the initial positions to obtain reference initial positions, and determining an initial range based on the reference initial positions;
determining a corresponding reference range based on each reference starting position, and obtaining a candidate range of interest points to be recommended according to the reference range and the starting range, wherein the candidate range of the interest points to be recommended comprises each interest point to be recommended;
acquiring an original interest point recommendation degree matrix corresponding to each interest point to be recommended;
performing matrix decomposition prediction based on the original interest point recommendation degree matrix to obtain a target interest point recommendation degree matrix, wherein the target interest point recommendation degree matrix comprises recommendation degrees of various objects to be recommended to various interest points to be recommended;
and determining a target recommendation interest point based on the interest point attribute information corresponding to each object to be recommended and the recommendation degree of each interest point to be recommended, and sending the target recommendation interest point to the terminal corresponding to each object to be recommended.
In one embodiment, calculating loss information of the original point of interest recommendation degree matrix and the reconstructed point of interest recommendation degree matrix includes:
calculating the error square sum of each recommendation degree in the original interest point recommendation degree matrix and the corresponding reconstruction recommendation degree in the reconstruction interest point recommendation degree matrix;
calculating first punishment item information corresponding to the first decomposition matrix based on a preset first punishment coefficient, and calculating second punishment item information corresponding to the second decomposition matrix based on a preset second punishment coefficient;
and obtaining the loss information based on the error square sum, the first penalty item information and the second penalty item information.
In one embodiment, calculating a sum of each candidate recommendation degree corresponding to each candidate interest point to obtain a sum of candidate recommendation degrees corresponding to each candidate interest point includes:
acquiring recommendation weights corresponding to the objects to be recommended, and performing weighted calculation on the candidate recommendation degrees corresponding to the objects to be recommended according to the recommendation weights to obtain the weighted candidate recommendation degrees;
and calculating the sum of the weighted candidate recommendation degrees of each candidate interest point to obtain the candidate recommendation degree sum of each candidate interest point.
In one embodiment, obtaining recommendation weights corresponding to objects to be recommended includes:
and obtaining voting information of each object to be recommended on the historical recommendation interest points, and calculating recommendation weights corresponding to the objects to be recommended based on the voting information.
A point of interest recommendation method, the method comprising:
responding to an interest point recommending event triggered by a target object to be recommended, and displaying an interest point recommending interface, wherein the interest point recommending interface displays an initial position corresponding to each object to be recommended, and each object to be recommended comprises the target object to be recommended;
responding to an interest point attribute information selection event triggered through an interest point recommendation interface, and displaying an interest point attribute information acquisition interface;
responding to an interest point attribute information confirmation event triggered by an interest point attribute information acquisition interface, acquiring target interest point attribute information corresponding to a target object to be recommended, and sending the target interest point attribute information to a server;
and displaying the target recommendation interest points returned by the server, wherein the target recommendation interest points are obtained by determining candidate ranges of the interest points to be recommended by the server according to initial positions corresponding to the objects to be recommended, determining the candidate interest points from the candidate ranges of the interest points to be recommended based on the attribute information of the interest points corresponding to the objects to be recommended and based on the candidate recommendation degrees corresponding to the candidate interest points, the candidate recommendation degrees corresponding to the candidate interest points are searched from a target interest point recommendation degree matrix, and the target interest point recommendation degree matrix is obtained by performing matrix decomposition prediction by using an original interest point recommendation degree matrix.
An apparatus for point of interest recommendation, the apparatus comprising:
the initial range determining module is used for acquiring initial positions and interest point attribute information corresponding to the objects to be recommended, screening the initial positions to obtain reference initial positions, and determining an initial range based on the reference initial positions;
a candidate range obtaining module, configured to determine a corresponding reference range based on each reference starting position, and obtain a candidate range of interest points to be recommended according to the reference range and the starting range, where the candidate range of interest points to be recommended includes each interest point to be recommended;
the original matrix obtaining module is used for obtaining an original interest point recommendation degree matrix corresponding to each interest point to be recommended;
the matrix decomposition module is used for carrying out matrix decomposition prediction based on the original interest point recommendation degree matrix to obtain a target interest point recommendation degree matrix, and the target interest point recommendation degree matrix comprises the recommendation degree of each object to be recommended to each interest point to be recommended;
and the interest point recommending module is used for determining a target recommending interest point based on the interest point attribute information corresponding to each object to be recommended and the recommendation degree of each interest point to be recommended and sending the target recommending interest point to the terminal corresponding to each object to be recommended.
A point of interest presentation apparatus, the apparatus comprising:
the recommendation interface display module is used for responding to an interest point recommendation event triggered by a target object to be recommended and displaying an interest point recommendation interface, wherein the interest point recommendation interface displays an initial position corresponding to each object to be recommended, and each object to be recommended comprises the target object to be recommended;
the information acquisition interface display module is used for responding to an interest point attribute information selection event triggered by the interest point recommendation interface and displaying an interest point attribute information acquisition interface;
the information sending module is used for responding to an interest point attribute information confirmation event triggered by the interest point attribute information acquisition interface, acquiring target interest point attribute information corresponding to a target object to be recommended and sending the target interest point attribute information to the server;
the interest point display module is used for displaying target recommendation interest points returned by the server, the target recommendation interest points are obtained by the server according to candidate recommendation degrees corresponding to the candidate interest points, the candidate recommendation degrees corresponding to the candidate interest points are searched from a target interest point recommendation degree matrix, and the target interest point recommendation degree matrix is obtained by performing matrix decomposition prediction by using an original interest point recommendation degree matrix.
A computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the following steps when executing the computer program:
acquiring initial positions and interest point attribute information corresponding to objects to be recommended, screening the initial positions to obtain reference initial positions, and determining an initial range based on the reference initial positions;
determining a corresponding reference range based on each reference starting position, and obtaining a candidate range of interest points to be recommended according to the reference range and the starting range, wherein the candidate range of the interest points to be recommended comprises each interest point to be recommended;
acquiring an original interest point recommendation degree matrix corresponding to each interest point to be recommended;
performing matrix decomposition prediction based on the original interest point recommendation degree matrix to obtain a target interest point recommendation degree matrix, wherein the target interest point recommendation degree matrix comprises recommendation degrees of various objects to be recommended to various interest points to be recommended;
and determining a target recommendation interest point based on the interest point attribute information corresponding to each object to be recommended and the recommendation degree of each interest point to be recommended, and sending the target recommendation interest point to the terminal corresponding to each object to be recommended.
A computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the following steps when executing the computer program:
responding to an interest point recommending event triggered by a target object to be recommended, and displaying an interest point recommending interface, wherein the interest point recommending interface displays an initial position corresponding to each object to be recommended, and each object to be recommended comprises the target object to be recommended;
responding to an interest point attribute information selection event triggered through an interest point recommendation interface, and displaying an interest point attribute information acquisition interface;
responding to an interest point attribute information confirmation event triggered by an interest point attribute information acquisition interface, acquiring target interest point attribute information corresponding to a target object to be recommended, and sending the target interest point attribute information to a server;
and displaying the target recommendation interest points returned by the server, wherein the target recommendation interest points are obtained by determining candidate ranges of the interest points to be recommended by the server according to initial positions corresponding to the objects to be recommended, determining the candidate interest points from the candidate ranges of the interest points to be recommended based on the attribute information of the interest points corresponding to the objects to be recommended and based on the candidate recommendation degrees corresponding to the candidate interest points, the candidate recommendation degrees corresponding to the candidate interest points are searched from a target interest point recommendation degree matrix, and the target interest point recommendation degree matrix is obtained by performing matrix decomposition prediction by using an original interest point recommendation degree matrix.
A computer-readable storage medium, on which a computer program is stored which, when executed by a processor, carries out the steps of:
acquiring initial positions and interest point attribute information corresponding to objects to be recommended, screening the initial positions to obtain reference initial positions, and determining an initial range based on the reference initial positions;
determining a corresponding reference range based on each reference starting position, and obtaining a candidate range of interest points to be recommended according to the reference range and the starting range, wherein the candidate range of the interest points to be recommended comprises each interest point to be recommended;
acquiring an original interest point recommendation degree matrix corresponding to each interest point to be recommended;
performing matrix decomposition prediction based on the original interest point recommendation degree matrix to obtain a target interest point recommendation degree matrix, wherein the target interest point recommendation degree matrix comprises recommendation degrees of various objects to be recommended to various interest points to be recommended;
and determining a target recommendation interest point based on the interest point attribute information corresponding to each object to be recommended and the recommendation degree of each interest point to be recommended, and sending the target recommendation interest point to the terminal corresponding to each object to be recommended.
A computer-readable storage medium, on which a computer program is stored which, when executed by a processor, carries out the steps of:
responding to an interest point recommending event triggered by a target object to be recommended, and displaying an interest point recommending interface, wherein the interest point recommending interface displays an initial position corresponding to each object to be recommended, and each object to be recommended comprises the target object to be recommended;
responding to an interest point attribute information selection event triggered through an interest point recommendation interface, and displaying an interest point attribute information acquisition interface;
responding to an interest point attribute information confirmation event triggered by an interest point attribute information acquisition interface, acquiring target interest point attribute information corresponding to a target object to be recommended, and sending the target interest point attribute information to a server;
displaying target recommendation interest points returned by the server, wherein the target recommendation interest points are obtained by determining candidate ranges of the interest points to be recommended by the server according to initial positions corresponding to the objects to be recommended, determining candidate interest points from the candidate ranges of the interest points to be recommended based on the attribute information of the interest points corresponding to the objects to be recommended and based on candidate recommendation degrees corresponding to the candidate interest points, the candidate recommendation degrees corresponding to the candidate interest points are searched from a target interest point recommendation degree matrix, and the target interest point recommendation degree matrix is obtained by performing matrix decomposition prediction by using an original interest point recommendation degree matrix
According to the interest point recommending method, the interest point recommending device, the computer equipment and the storage medium, the initial position and the interest point attribute information corresponding to each object to be recommended are obtained, and the candidate range of the interest point to be recommended and each interest point to be recommended are determined according to each initial position. Then obtaining an original interest point recommendation degree matrix corresponding to each object to be recommended, performing matrix decomposition prediction by using the original interest point recommendation degree matrix to obtain a target interest point recommendation degree matrix, then determining a target recommendation interest point based on the interest point attribute information corresponding to each object to be recommended and the recommendation degree of each interest point to be recommended, sending the target recommendation interest point to a terminal corresponding to each object to be recommended, namely obtaining the recommendation degree corresponding to the interest point to be recommended in a candidate range of the interest point to be recommended through matrix decomposition, ensuring the accuracy of the obtained recommendation degree, then determining candidate interest points from each interest point to be recommended according to the interest point attribute information corresponding to each object to be recommended, determining the target recommendation interest points from the candidate interest points based on the candidate recommendation degree of the candidate interest points, and thus improving the accuracy of the target recommendation interest points obtained when a plurality of objects to be recommended are recommended, and then recommending the target recommendation interest points to each object to be recommended, so that the accuracy of recommending the interest points of a plurality of objects to be recommended is improved.
According to the method, the device, the computer equipment and the storage medium for displaying the interest points, the terminal responds to the interest point recommending event triggered by the target object to be recommended, displays the interest point recommending interface, responds to the interest point attribute information selecting event triggered by the interest point recommending interface, displays the interest point attribute information acquiring interface, responds to the interest point attribute information confirming event triggered by the interest point attribute information acquiring interface, acquires the target interest point attribute information corresponding to the target object to be recommended, sends the target interest point attribute information to the server, displays the target recommending interest points returned by the server, so that the target object to be recommended can acquire the accurate target recommending interest points, the target recommending interest points displayed by the terminal are more accurate, the terminal can display the initial positions of the objects to be recommended, namely, the common content can be synchronously displayed, and then the terminal can acquire the interest point attribute information corresponding to the target object to be recommended through the interest point attribute information acquisition interface, namely the terminal can also display personalized content, so that the terminal is convenient for a user to use.
Drawings
FIG. 1 is a diagram of an application environment of a point of interest recommendation method in one embodiment;
FIG. 2 is a flowchart illustrating a point of interest recommendation method according to an embodiment;
FIG. 3 is a flow diagram illustrating the process of determining a starting range in one embodiment;
FIG. 4 is a schematic illustration of the starting range in one embodiment;
FIG. 5 is a schematic flow chart illustrating obtaining a first partial reference starting position according to one embodiment;
FIG. 6 is a diagram of an original point of interest recommendation matrix in an exemplary embodiment;
FIG. 7 is a flowchart illustrating obtaining a recommendation degree matrix for a target point of interest in one embodiment;
FIG. 8 is a flow diagram illustrating obtaining loss information in one embodiment;
FIG. 9 is a schematic diagram illustrating a process for obtaining a target recommended point of interest in one embodiment;
FIG. 10 is a flow diagram that illustrates the determination of a final recommended point of interest, in one embodiment;
FIG. 11 is a flow diagram that illustrates the determination of model recommended points of interest, in one embodiment;
FIG. 12 is a flowchart illustrating a method for showing points of interest in one embodiment;
FIG. 13 is a diagram of a point of interest recommendation interface in accordance with an illustrative embodiment;
FIG. 14 is a diagram of a point of interest attribute information acquisition interface in a particular embodiment;
FIG. 15 is a flowchart illustrating a point of interest recommendation method in an exemplary embodiment;
FIG. 16 is a flowchart illustrating a point of interest recommendation method in accordance with another exemplary embodiment;
FIG. 17 is a block diagram showing the structure of a point of interest recommendation apparatus in one embodiment;
FIG. 18 is a block diagram of a point of interest presentation apparatus in one embodiment;
FIG. 19 is a diagram showing an internal structure of a computer device in one embodiment;
fig. 20 is an internal structural view of a computer device in another embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
The interest point recommendation method provided by the application can be applied to the application environment shown in fig. 1. Each terminal includes a terminal 102, a terminal 104, and a terminal 106, which communicate with the server 104 via a network. The server 104 obtains the initial position and the interest point attribute information corresponding to each object to be recommended through each terminal, screens the initial positions to obtain reference initial positions, and determines an initial range based on the reference initial positions. The server 104 determines a corresponding reference range based on each reference starting position, and obtains a candidate range of the interest points to be recommended according to the reference range and the starting range, wherein the candidate range of the interest points to be recommended includes each interest point to be recommended. The server 104 acquires an original interest point recommendation degree matrix corresponding to each interest point to be recommended. The server 104 performs matrix decomposition prediction based on the original interest point recommendation degree matrix to obtain a target interest point recommendation degree matrix, wherein the target interest point recommendation degree matrix includes recommendation degrees of the objects to be recommended to the interest points to be recommended. And determining a target recommendation interest point based on the interest point attribute information corresponding to each object to be recommended and the recommendation degree of each interest point to be recommended, and sending the target recommendation interest point to the terminal corresponding to each object to be recommended. The terminal can be but not limited to various personal computers, notebook computers, smart phones, tablet computers and portable wearable devices, the server can be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, and can also be a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, a CDN (content distribution network), big data platforms and artificial intelligence platforms. The terminal may be, but is not limited to, a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, and the like. The terminal and the server may be directly or indirectly connected through wired or wireless communication, and the application is not limited herein.
In one embodiment, as shown in fig. 2, a method for recommending a point of interest is provided, which is described by taking the method as an example for being applied to the server in fig. 1, and includes the following steps:
step 202, obtaining the initial position and the interest point attribute information corresponding to each object to be recommended, screening from each initial position to obtain each reference initial position, and determining an initial range based on each reference initial position.
The object to be recommended refers to a user object needing to be subjected to interest point recommendation. The point of interest attribute information refers to specific requirement information of an object to be recommended for a point of interest to be recommended, the point of interest attribute information may include specific requirement information of at least one category, the category may be a catering category, a movie category, a shopping category, a point of interest category, and the like, and the specific requirement information of the category refers to requirement information of basic attribute information of the category. For example, the specific requirement information of the food and beverage category may include pungency information, cuisine information, price information, and the like. For example, the specific requirement information of the movie category may be movie type information, movie time length information, movie price information, and the like. And each object to be recommended has corresponding interest point attribute information. The starting position refers to a starting position corresponding to the object to be recommended, the object to be recommended starts from the starting position to the recommended interest point, and the starting position can be represented in a coordinate form. The start range refers to a range of convex polygonal areas, i.e., convex hulls, that can completely encompass each start position. The reference start position is a start position of a vertex in the start range.
Specifically, the server obtains the initial position and the interest point attribute information corresponding to each object to be recommended from the terminal corresponding to each object to be recommended. And then, each initial position is used for convex hull calculation, namely, screening is carried out from each initial position to obtain a vertex on the convex hull, the vertex on the convex hull is used as each reference initial position, and then each reference initial position is connected to obtain the convex hull, namely, the initial range. In an embodiment, the server may obtain the stored start position and interest point attribute information corresponding to each object to be recommended from the database. In an embodiment, the server may also obtain the start position and the attribute information of the point of interest corresponding to each object to be recommended, which is pushed by a third-party server, where the third-party server is a service party that needs to perform point of interest recommendation, and the service party invokes a point of interest recommendation method in the server to perform recommendation.
And 204, determining a corresponding reference range based on each reference starting position, and obtaining a candidate range of the interest points to be recommended according to the reference range and the starting range, wherein the candidate range of the interest points to be recommended comprises each interest point to be recommended.
The reference range refers to a range of an area determined centering on the reference start position. The candidate range of the interest points to be recommended refers to a range for selecting the interest points to recommend. The interest points to be recommended refer to the interest points needing to be recommended after being screened. The interest points to be recommended comprise all the interest points in the candidate range of the interest points to be recommended.
Specifically, the server may determine the circular region by taking each reference start position as a center according to a preset radius, so as to obtain a reference range corresponding to each reference start position. And then combining the reference ranges and the initial ranges to obtain a candidate range of the interest points to be recommended, wherein the candidate range of the interest points to be recommended comprises the interest points to be recommended. In an embodiment, the server may also determine other planar geometric shape areas with each reference start position as a center, to obtain a reference range corresponding to each reference start position. For example, a square area, a triangular area, a rectangular area, or the like may be determined with each reference start position as a center, and a reference range corresponding to each reference start position may be obtained. And then merging according to the obtained reference ranges and the initial ranges to obtain the candidate range of the interest point to be recommended.
And step 206, acquiring an original interest point recommendation degree matrix corresponding to each interest point to be recommended.
The original interest point recommendation degree matrix is a recommendation degree matrix of each object to be recommended to each interest point to be recommended and each historical interest point, and is a sparse matrix, the historical interest points in the original interest point recommendation degree matrix have corresponding historical recommendation degrees, and each interest point to be recommended has no corresponding recommendation degree, so that the recommendation degree of each interest point to be recommended can be initialized to be zero. The historical interest points refer to interest points with historical recommendation degrees corresponding to the objects to be recommended, that is, the interest points that the objects to be recommended have already passed. The historical recommendation degree refers to the recommendation degree of each object to be recommended to the historical interest point, and can be obtained according to evaluation information of each object to be recommended to the historical interest point.
Specifically, the server may obtain the historical recommendation degrees of the historical interest points corresponding to the objects to be recommended from the database, and then perform matrix building by using the interest points to be recommended, the historical interest points, and the historical recommendation degrees corresponding to the objects to be recommended corresponding to the historical interest points, to obtain an original interest point recommendation degree matrix. For example, when there are m objects to be recommended, and when there are n interest points to be recommended and n historical interest points in total, the obtained original interest point recommendation degree matrix is Rm×nAnd m and n are positive integers.
And 208, performing matrix decomposition prediction based on the original interest point recommendation degree matrix to obtain a target interest point recommendation degree matrix, wherein the target interest point recommendation degree matrix comprises the recommendation degree of each object to be recommended to each interest point to be recommended.
The target interest point recommendation degree matrix is an interest point recommendation degree matrix obtained through matrix decomposition prediction, and each object to be recommended in the target interest point recommendation degree matrix has a predicted recommendation degree for each interest point to be recommended.
Specifically, the server performs matrix decomposition on the original interest point recommendation degree matrix to obtain a decomposed matrix, then performs iterative optimization on the decomposed matrix to obtain an optimized decomposition matrix, and then merges the optimized decomposition matrices to obtain a target interest point recommendation degree matrix, wherein the target interest point recommendation degree matrix comprises recommendation degrees of objects to be recommended to the interest points to be recommended.
Step 210, determining a target recommendation interest point based on the interest point attribute information corresponding to each object to be recommended and the recommendation degree of each interest point to be recommended, and sending the target recommendation interest point to a terminal corresponding to each object to be recommended.
The target recommendation interest points are interest points which recommend to each object to be recommended and comprise at least one candidate interest point.
Specifically, the server determines candidate interest points from the interest points to be recommended based on the interest point attribute information corresponding to each object to be recommended, searches for candidate recommendation degrees corresponding to the candidate interest points from the recommendation degrees of the interest points to be recommended, and determines target interest points from the candidate interest points based on the candidate recommendation degrees. And calculating the sum of each candidate recommendation degree of each candidate interest point, wherein the candidate interest points correspond to the recommendation degrees of the objects to be recommended. Then, calculating the average of the sum of the candidate recommendation degrees to obtain the average recommendation degree corresponding to each candidate interest point, then sequentially ordering the candidate interest points according to the average recommendation degree corresponding to each candidate interest point to obtain a candidate interest point sequence, and then sequentially selecting the target recommendation interest points from the candidate interest point sequence. And then sending the selected target recommendation interest points to terminals corresponding to the objects to be recommended, and displaying the terminals corresponding to the objects to be recommended when the terminals receive the target recommendation interest points.
According to the interest point recommendation method, the starting positions and the interest point attribute information corresponding to the objects to be recommended are obtained, and the candidate range of the interest points to be recommended and the interest points to be recommended are determined according to the starting positions. Then, establishing an original interest point recommendation degree matrix by using the historical recommendation degree of the historical interest points and each interest point to be recommended, performing matrix decomposition prediction by using the original interest point recommendation degree matrix to obtain a target interest point recommendation degree matrix, then determining candidate interest points from each interest point to be recommended based on the interest point attribute information corresponding to each object to be recommended, searching for candidate recommendation degrees corresponding to the candidate interest points from the recommendation degrees of the interest points to be recommended, then determining target recommendation interest points from the candidate interest points based on the candidate recommendation degrees, sending the target recommendation interest points to terminals corresponding to each object to be recommended, namely obtaining recommendation degrees corresponding to the interest points to be recommended in a candidate range of the interest points to be recommended through matrix decomposition, ensuring the accuracy of the obtained recommendation degrees, and then determining the candidate interest points from each interest point to be recommended according to the interest point attribute information corresponding to each object to be recommended, the target recommendation interest points are determined from the candidate interest points based on the candidate recommendation degrees of the candidate interest points, so that the accuracy of the target recommendation interest points obtained when a plurality of objects to be recommended are recommended can be improved, then the target recommendation interest points are recommended to all the objects to be recommended, and the accuracy of the interest point recommendation of the plurality of objects to be recommended is improved.
In one embodiment, as shown in fig. 3, step 202, obtaining each reference starting position by filtering from each starting position, and determining a starting range based on each reference starting position, includes:
step 302, a first reference starting position and a second reference starting position are determined based on the position coordinates corresponding to the starting positions.
The position coordinates are coordinates for representing the starting position, and may be two-dimensional plane coordinates. The first reference starting position refers to a starting position whose abscissa is the smallest in the position coordinates. The second reference starting position refers to a starting position having the largest abscissa among the position coordinates.
Specifically, the server sorts the abscissa of the position coordinates of the initial positions from small to large to obtain a sorting result of the initial positions, selects the initial position corresponding to the minimum abscissa as a first reference initial position, and selects the initial position corresponding to the maximum abscissa as a second reference initial position. In one embodiment, the position coordinates may also be longitude and latitude coordinates, and the first reference starting position and the second reference starting position are determined according to the size of the longitude and latitude coordinates.
Step 304, determining a first straight line based on the first reference starting position and the second reference starting position, and dividing each starting position based on the first straight line to obtain a first part starting position and a second part starting position.
Wherein the first straight line is a straight line connecting the first reference starting position and the second reference starting position for dividing each starting position into two parts. The first part start position is a start position of an upper part divided by a first straight line. The second portion start position refers to a start position of a lower portion obtained by dividing using the first straight line.
Specifically, the server determines a first straight line according to a first reference starting position and a second reference starting position, and divides each starting position by using the first straight line to obtain each divided starting position of different parts, namely a first part starting position and a second part starting position.
Step 306, filtering from the first part starting positions to obtain the first part reference starting positions, and filtering from the second part starting positions to obtain the second part reference starting positions.
The reference starting position of the first part refers to the starting position of the convex hull vertex in the first part. The second portion reference starting position refers to the starting position of the convex hull vertex in the second portion.
Specifically, the server recursively calculates respective first-section reference start positions contained in the first-section start positions. The server can calculate the area of a triangle formed by each initial position in the first division initial position and the first straight line, the initial position with the largest area is used as a first part reference initial position, then the initial position with the largest area is connected with the first reference initial position to obtain a straight line, the first part initial position is divided by using the straight line to obtain a divided result, then the divided reference initial positions are calculated in a recursion mode, when the initial position which can not form the triangle with the straight line after the division is not available, the recursion is finished, and the reference initial positions are obtained according to the recursion calculation to obtain the first part reference initial positions. While the server recursively calculates respective second part reference start positions contained in the first part start positions. The server can calculate the area of a triangle formed by each initial position in the second division initial position and the first straight line, the initial position with the largest area is used as a second division reference initial position, then the initial position with the largest area is connected with the first reference initial position to obtain a straight line, the second division initial position is divided by using the straight line to obtain a divided result, then the divided reference initial positions are calculated in a recursion mode, when the initial position which can not form the triangle with the straight line after the division is not available, the recursion is finished, and the reference initial positions are obtained according to the recursion calculation to obtain the reference initial positions of each second division reference initial position.
Step 308, obtaining each reference starting position based on the first reference starting position, the second reference starting position, each first partial reference starting position and each second partial reference starting position, and determining a starting range based on each reference starting position.
Specifically, the server directly takes the obtained first reference starting position, the second reference starting position, each first part reference starting position and each second part reference starting position as each reference starting position, and then takes an area range obtained by sequentially connecting the reference starting positions as a starting range. In a specific embodiment, as shown in fig. 4, a schematic diagram of the starting range is shown, wherein black dots represent each starting position, and the area of the convex polygon represents the starting range.
In one embodiment, as shown in fig. 5, step 306 of filtering from the starting positions of the first portions to obtain the reference starting positions of the respective first portions includes:
in step 502, a vertical distance between each starting position in the first part starting position and the first straight line is calculated, and a third reference starting position is determined from the first part starting position based on the vertical distance.
Wherein the third reference starting position refers to a reference starting position determined from the first part starting position, i.e. the vertex of the convex hull.
Specifically, the server calculates a perpendicular distance from the first straight line, i.e., a point-to-straight line distance, from the position coordinates of each of the start positions in the first partial start positions. And then comparing the vertical centers corresponding to the initial positions to obtain the initial position corresponding to the maximum vertical distance, and taking the initial position corresponding to the maximum vertical distance as a third reference initial position.
Step 504, determining a second straight line based on the third reference starting position and the first reference starting position, dividing the first part starting position based on the second straight line to obtain a third part starting position, determining a third straight line based on the third reference starting position and the second reference starting position, and dividing the first part starting position based on the third straight line to obtain a fourth part starting position.
Wherein the second straight line is a straight line determined according to the third reference starting position and the first reference starting position. The third straight line is a straight line determined according to the third reference starting position and the second reference starting position, and the third part starting position is the starting position of the left part obtained by dividing the first part starting position according to the second straight line. The fourth starting position is the starting position of the right part obtained by dividing the first starting position according to the third straight line.
Specifically, the server determines a second straight line according to a third reference starting position and the first reference starting position, divides the first part starting position by using the second straight line, determines a third part starting position from the division result, determines a third straight line by using the third reference starting position and the second reference starting position, divides the first part starting position by using the third straight line, and determines a fourth part starting position from the division result.
Step 506, taking the third part starting position and the fourth part starting position as the first part starting position respectively, taking the second straight line and the third straight line as the first straight line respectively, calculating the vertical distance between each starting position in the first part starting position and the first straight line, and determining a third reference starting position from the first part starting position based on the vertical distance.
Specifically, the server performs recursive computation, the server may first use the third start position as the first start position, use the second straight line as the first straight line, return to computing the vertical distance between each start position in the first start position and the first straight line, perform the step of determining a third reference start position from the first start position based on the vertical distance, obtain each reference start position in the third start position until the third start position is not included in the third start position, then use the fourth start position as the first start position, use the third straight line as the first straight line, return to computing the vertical distance between each start position in the first start position and the first straight line, perform the step of determining a third reference start position from the first start position based on the vertical distance until the fourth start position is not included in the fourth start position, and obtaining each reference starting position in the fourth part starting position, and obtaining each first part reference starting position by the server according to each reference starting position in the third part starting position and each reference starting position in the fourth part starting position.
In one embodiment, the server may filter the second part starting positions to obtain the reference starting positions of the second parts, specifically, the server calculates vertical distances between the first line and the respective starting positions of the second part starting positions, and determines a fifth reference starting position from the second part starting positions based on the vertical distances. Determining a fifth straight line based on the fifth reference starting position and the first reference starting position, dividing the second part starting position based on the fifth straight line to obtain a fifth part starting position, determining a sixth straight line based on the fifth reference starting position and the second reference starting position, and dividing the second part starting position based on the sixth straight line to obtain a sixth part starting position. And respectively taking the starting position of the fifth part and the starting position of the sixth part as the starting position of the first part, respectively taking the fifth straight line and the sixth straight line as the first straight line, returning to calculate the vertical distance between each starting position in the starting position of the first part and the first straight line, and determining a fifth reference starting position from the starting position of the first part based on the vertical distance.
In the above embodiment, the reference starting positions of the respective first portions are obtained by filtering from the starting positions of the first portions, and the reference starting positions of the respective second portions are obtained by filtering from the starting positions of the second portions.
In one embodiment, the server may calculate the start range based on each start position by using a Graham scanning algorithm, that is, by determining the start position with the smallest ordinate, and then sequentially searching the vertices on the convex hull from the start position with the smallest ordinate in the counterclockwise direction from the start position with the smallest ordinate to obtain each reference start position, and then obtaining the start range according to each start position. In one embodiment, the server may also calculate the starting range using a wrapping algorithm, a fast wrapping algorithm, or the like based on the respective starting locations.
In one embodiment, step 204, determining a corresponding reference range based on each reference starting position, and obtaining a candidate range of interest points to be recommended according to the reference range and the starting range includes:
and determining each reference range based on each reference starting position and the preset circle radius, and combining each reference range and the starting range to obtain the candidate range of the interest point to be recommended.
The preset circle radius refers to a preset circle radius for determining a reference range, and can be set according to business requirements.
Specifically, the server uses each reference starting position as a circle center, and performs area circle range division according to a preset circle radius to obtain each reference range. And then the server combines each reference range with the initial range, namely combines the ranges which are not repeated with the initial range in each reference range, and takes the combined area range as the candidate range of the interest point to be recommended.
In the embodiment, the reference ranges and the initial range are combined to obtain the candidate range of the interest point to be recommended, the reference ranges corresponding to the objects to be recommended are comprehensively considered, the accuracy of the candidate range of the interest point to be recommended is improved,
in an embodiment, in step 206, obtaining an original interest point recommendation degree matrix corresponding to each interest point to be recommended includes:
obtaining historical recommendation degrees of historical interest points corresponding to the objects to be recommended, obtaining matrix column attributes based on the interest points to be recommended and the historical interest points, obtaining matrix row attributes based on the objects to be recommended, taking the historical recommendation degrees of the historical interest points as matrix values, and obtaining an original interest point recommendation degree matrix based on the matrix column attributes, the matrix row attributes and the matrix values.
The matrix column attribute refers to an attribute represented by a column in a matrix. Matrix row attributes refer to attributes characterized by rows in a matrix. Matrix values refer to specific values in a matrix.
Specifically, the server takes each interest point to be recommended and each historical interest point as matrix column attributes, takes each object to be recommended as a matrix row attribute, then takes the historical recommendation degree of the historical interest point as a matrix value, and obtains an original interest point recommendation degree matrix based on the matrix column attributes, the matrix row attributes and the matrix values. In a specific embodiment, as shown in fig. 6, a schematic diagram of an obtained original interest point recommendation degree matrix is shown, where the number of interest points is a, the number of historical interest points is B, the total number of the interest points and the historical interest points is n, m objects to be recommended are provided, "0" indicates that the recommendation degree of the interest points does not exist and is an initial value, a and B indicate the historical recommendation degree corresponding to the historical interest points, and different objects to be recommended have different historical recommendation degrees.
In the embodiment, the matrix column attribute is obtained according to each interest point to be recommended and the historical interest points, the matrix row attribute is obtained based on each object to be recommended, the historical recommendation degree of the historical interest points is used as the matrix value, and the original interest point recommendation degree matrix is established, so that the subsequent use can be facilitated, and the efficiency is improved.
In one embodiment, as shown in fig. 7, in step 208, performing matrix decomposition prediction based on the original interest point recommendation degree matrix to obtain a target interest point recommendation degree matrix, including:
step 702, performing matrix decomposition on the original interest point recommendation degree matrix according to a preset decomposition relationship to obtain a first decomposition matrix and a second decomposition matrix, wherein the first decomposition matrix is used for representing the recommendation degree relationship between each object to be recommended and each attribute in the interest point attribute information, and the second decomposition matrix is used for representing the weight relationship between each attribute in the interest point attribute information and the interest point;
the preset decomposition relationship refers to a preset relationship for decomposing the matrix. The first decomposition matrix is a matrix which is obtained by matrix decomposition of the original interest point recommendation degree matrix and is used for representing recommendation degree relations between each object to be recommended and each attribute in the interest point attribute information. The second decomposition matrix is obtained by matrix decomposition of the original interest point recommendation degree matrix and is used for representing the weight relation between each attribute in the interest point attribute information and the interest point. The recommendation degree relation refers to the recommendation degree of the object to be recommended to each attribute in the attribute information of the interest point. The weight relationship refers to the weight of each attribute in the attribute information of the interest point relative to the interest point.
Specifically, the server splits the original interest point recommendation degree matrix according to a preset pre-solution relationship to obtain a first decomposition matrix and a second decomposition matrix, where matrix values in the first decomposition matrix and the second decomposition matrix may be random initial values. For example, the splitting may be performed according to the following formula (1), and the split first split matrix P is obtainedm×kAnd a second decomposition matrix Qk×n。
Rm×n≈Pm×k×Qk×nFormula (1)
Wherein R ism×nThe original interest point recommendation degree matrix is represented and can be a dilution 0-1 matrix. Pm×kAnd the matrix represents the recommendation degree relation between the m objects to be recommended and the k attributes in the interest point attribute information. Qk×nRepresenting the weight relationship between k attributes and n interest points in the interest point attribute informationA matrix of systems. And calculating a first decomposition matrix and a second decomposition matrix by using iterative optimization of matrix decomposition, so that the first decomposition matrix and the second decomposition matrix approach to the original interest point recommendation degree matrix as much as possible.
Step 704, calculating a product of the first decomposition matrix and the second decomposition matrix to obtain a reconstructed interest point recommendation degree matrix.
The reconstructed interest point recommendation degree matrix refers to a reconstructed interest point recommendation degree matrix, and the rows and columns of the reconstructed interest point recommendation degree matrix are consistent with the original interest point recommendation degree matrix.
Specifically, the server iteratively optimizes the first decomposition matrix and the second decomposition matrix, namely, the first decomposition matrix and the second decomposition matrix are used as implicit feedback, and the product of the first decomposition matrix and the second decomposition matrix is calculated, namely, matrix multiplication calculation is performed, so that a reconstructed interest point recommendation degree matrix is obtained.
Step 706, calculating loss information of the original interest point recommendation degree matrix and the reconstructed interest point recommendation degree matrix, and updating the first decomposition matrix and the second decomposition matrix based on the loss information.
The loss information refers to an error between the recommendation degree in the reconstructed interest point recommendation degree matrix and the recommendation degree in the corresponding original interest point recommendation degree matrix.
Specifically, the server calculates the errors between the recommendation degrees in the original interest point recommendation degree matrix and the corresponding recommendation degrees in the reconstructed interest point recommendation degree matrix by using a square error loss function to obtain loss information, and then updates each matrix value in the first decomposition matrix and each matrix value in the second decomposition matrix by using the loss information.
And 708, calculating the product of the first decomposition matrix and the second decomposition matrix, and iteratively executing the step of obtaining the reconstructed interest point recommendation degree matrix until the target first decomposition matrix and the target second decomposition matrix are obtained when the loss information meets the preset loss condition.
Wherein, the target first decomposition matrix refers to the first decomposition matrix after the iterative optimization. The target second decomposition matrix refers to the second decomposition matrix after the iterative optimization is completed.
Specifically, the server determines whether the loss information meets a preset loss condition, where the preset loss condition may be that the loss information is minimum, that is, an error between the recommendation degree in the reconstructed interest point recommendation degree matrix and the recommendation degree in the corresponding original interest point recommendation degree matrix is minimum. When the loss information does not meet the preset loss condition, the server returns to step 706, that is, the step of calculating the product of the first decomposition matrix and the second decomposition matrix is returned to obtain the reconstructed interest point recommendation degree matrix is iteratively executed until the target first decomposition matrix and the target second decomposition matrix are obtained when the loss information meets the preset loss condition.
And 710, calculating the product of the target first decomposition matrix and the target second decomposition matrix to obtain a target interest point recommendation degree matrix.
Specifically, the server performs matrix multiplication calculation on the target first decomposition matrix and the target second decomposition matrix to obtain a target interest point recommendation degree matrix. For example, the target interest point recommendation degree matrix may be obtained by calculation using the following formula (2).
Wherein,representing the target first decomposition matrix.Representing the target second decomposition matrix.And representing a recommendation degree matrix of the target interest points.
In a specific embodiment, the target interest point recommendation degree matrix may be calculated by using ALS (Alternating Least Square).
In the above embodiment, the first decomposition matrix and the second decomposition matrix are iteratively optimized by using a square error loss function, so that a target first decomposition matrix and a target second decomposition matrix are obtained, and then a product of the target first decomposition matrix and the target second decomposition matrix is calculated to obtain a target interest point recommendation degree matrix, so that the accuracy of the obtained target interest point recommendation degree matrix is improved.
In one embodiment, as shown in fig. 8, step 706, calculating loss information of the original point of interest recommendation matrix and the reconstructed point of interest recommendation matrix includes:
and step 802, calculating the sum of the squares of errors of each recommendation degree in the original interest point recommendation degree matrix and the corresponding reconstruction recommendation degree in the reconstruction interest point recommendation degree matrix.
Specifically, the server obtains the historical recommendation degrees corresponding to the historical interest points in the original interest point recommendation degree matrix, and obtains the reconstruction recommendation degrees corresponding to the historical interest points in the reconstruction interest point recommendation degree matrix. And then calculating the square of the difference between each historical recommendation degree and the corresponding reconstruction recommendation degree to obtain the error between each historical recommendation degree and the corresponding reconstruction recommendation degree, and then calculating the sum of all the errors to obtain the sum of the squares of the errors.
Step 802, calculating first penalty item information corresponding to the first decomposition matrix based on a preset first penalty coefficient, and calculating second penalty item information corresponding to the second decomposition matrix based on a preset second penalty coefficient.
The preset first penalty coefficient refers to a preset penalty coefficient for a regular term of the first decomposition matrix. The preset second penalty coefficient refers to a preset penalty coefficient for a regular term of the second decomposition matrix. The first penalty item information is information obtained by calculation by using a preset first penalty coefficient and a regular item corresponding to the first decomposition matrix. The second penalty item information is information obtained by calculation by using a preset second penalty coefficient and a regular item corresponding to the second decomposition matrix.
Specifically, the server calculates a matrix norm by using the first decomposition matrix to obtain a value of a regular term corresponding to the first decomposition matrix, and then calculates a product of the value of the regular term and a preset first penalty coefficient to obtain first penalty term information. Similarly, the server calculates the matrix norm corresponding to the first decomposition matrix to obtain the value of the regular term corresponding to the second decomposition matrix, and then calculates the product of the value of the regular term and the preset second penalty coefficient to obtain the second penalty term information.
And step 802, obtaining loss information based on the error square sum, the first penalty item information and the second penalty item information.
Specifically, the server calculates the sum of the squared errors and the sum of the first penalty item information and the second penalty item information to obtain loss information. In a specific embodiment, the server may calculate the loss information using equation (3) as shown below.
Where loss represents loss information. r isuiAnd representing the historical recommendation degree corresponding to each historical interest point in the original interest point recommendation degree matrix.And representing the reconstruction recommendation degree corresponding to the historical recommendation degree in the reconstruction interest point recommendation degree matrix. p is a radical ofu,kRepresenting matrix values, q, in a first decomposition matrix of an objectk.iRepresenting the target second decomposition matrix. Lambda [ alpha ]pRepresenting a preset first penalty factor. Lambda [ alpha ]qRepresenting a preset second penalty factor. Lambda [ alpha ]p||pu||2Indicating first penalty item information, λq||qi||2Indicating second penalty item information. u belongs to {0,1,2,. eta., m }, and m is the number of objects to be recommended. i ∈ {0,1,2,. and n }, where n is the sum of the historical interest points and the number of interest points.
In the embodiment, the loss information is obtained by calculating the first penalty item information and the second penalty item information and then using the error sum of squares, the first penalty item information and the second penalty item information, so that the generalization capability is improved, the iterative optimization overfitting is prevented, and the obtained loss information is more accurate.
In one embodiment, as shown in fig. 9, the step 212 of determining a target recommendation interest point based on the interest point attribute information and the recommendation degree of each to-be-recommended interest point includes:
step 902, determining candidate interest points from the interest points to be recommended based on the interest point attribute information corresponding to each object to be recommended, and searching for candidate recommendation degrees corresponding to the candidate interest points from the recommendation degrees of the interest points to be recommended.
The candidate interest points are obtained by screening the interest points to be recommended by using the interest point attribute information corresponding to the objects to be recommended. The candidate recommendation degree refers to the recommendation degree of the candidate interest point.
Specifically, the server may obtain attribute information of each to-be-recommended interest point, match the attribute information of each to-be-recommended interest point with the interest point attribute information corresponding to each to-be-recommended object, and when there is interest point attribute information that matches consistently, take the to-be-recommended interest point corresponding to the interest point attribute information that matches consistently as a candidate interest point. The matching may be consistent, which may mean that the categories of the attribute information of the points of interest are the same, and the specific requirements of the categories are the same. For example, the category in the attribute information of the interest point to be recommended is a catering category, the specific requirement information is that the pungency degree is slightly spicy, the cuisine is Hunan cuisine, and the like, and then the attribute information of the interest point corresponding to the object to be recommended is the same as the attribute information of the interest point to be recommended, and the matching is consistent. And then the server searches the recommendation degrees corresponding to the interest points to be recommended and consistent with the candidate interest points in the interest points to be recommended of the target interest point recommendation degree matrix, and takes the recommendation degrees corresponding to the consistent interest points to be recommended as the candidate recommendation degrees corresponding to the candidate interest points. The candidate recommendation degree comprises the recommendation degree of each object to be recommended corresponding to the candidate interest point.
And 904, acquiring the number of objects to be recommended, and calculating the sum of each candidate recommendation degree corresponding to each candidate interest point to obtain the sum of the candidate recommendation degrees corresponding to each candidate interest point.
The number of objects to be recommended refers to the total number of each object to be recommended, and may be, for example, the total number of users in a party. The sum of the candidate recommendation degrees refers to the sum of all candidate recommendation degrees corresponding to the candidate interest points.
Specifically, the server may count the number of objects to be recommended obtained by each object to be recommended. And then calculating the sum of the candidate recommendation degrees corresponding to the objects to be recommended corresponding to each candidate interest point to obtain the sum of the candidate recommendation degrees corresponding to each candidate interest point. For example, when the number of the objects to be recommended is 5, each candidate interest point corresponds to 5 candidate recommendation degrees, and the sum of the 5 candidate recommendation degrees is calculated to obtain the sum of the candidate recommendation degrees corresponding to the candidate interest points.
Step 906, performing average calculation based on the sum of the candidate recommendation degrees and the number of the objects to be recommended to obtain the average candidate recommendation degree corresponding to each candidate interest point.
Specifically, the server calculates the ratio of the sum of the candidate recommendation degrees to the number of the objects to be recommended, that is, performs average calculation to obtain the average candidate recommendation degree corresponding to each candidate interest point.
Step 908, the average candidate recommendation degrees corresponding to the candidate interest points are ranked, and a preset number of candidate interest points are selected as target recommendation interest points according to the ranking result.
Specifically, the server sorts the candidate interest points in sequence from large to small according to the average candidate recommendation degree to obtain the sorted candidate interest points, and then a preset number of candidate interest points can be selected from the sorted candidate interest points as target recommendation interest points, for example, the candidate interest points at the top three of the sorting can be selected as target recommendation interest points.
In the embodiment, the average candidate recommendation degree is calculated, and the candidate interest points are selected as the target recommendation interest points according to the average candidate recommendation degree, so that the selected target recommendation interest points are accurate, and the requirements of each object to be recommended can be met.
In one embodiment, step 904, calculating a sum of candidate recommendation degrees corresponding to each candidate interest point to obtain a sum of candidate recommendation degrees corresponding to each candidate interest point, includes the steps of:
acquiring recommendation weights corresponding to the objects to be recommended, and performing weighted calculation on the candidate recommendation degrees corresponding to the objects to be recommended according to the recommendation weights to obtain the weighted candidate recommendation degrees; and calculating the sum of the weighted candidate recommendation degrees of each candidate interest point to obtain the candidate recommendation degree sum of each candidate interest point.
The recommendation weight is used for representing the weight of the object to be recommended. Different objects to be recommended have different recommendation weights.
Specifically, the server obtains a recommendation weight corresponding to each object to be recommended, where the recommendation weight may be preset, or the server may calculate a weight of each object to be recommended in advance according to existing past data, where the existing past data may be behavior data of a history selection interest point corresponding to each object to be recommended. Then the server uses the recommendation weight to perform weighted calculation on the candidate recommendation degrees corresponding to the objects to be recommended to obtain the weighted candidate recommendation degrees; and then the server calculates the sum of the weighted candidate recommendation degrees corresponding to each candidate interest point to obtain the sum of the candidate recommendation degrees corresponding to each candidate interest point. In the embodiment, the sum of the candidate recommendation degrees is calculated after the candidate recommendation degrees are weighted and calculated according to the recommendation weight, so that the obtained sum of the candidate recommendation degrees is more accurate.
In one embodiment, obtaining recommendation weights corresponding to objects to be recommended includes the steps of:
and obtaining voting information of each object to be recommended on the historical recommendation interest points, and calculating recommendation weights corresponding to the objects to be recommended based on the voting information.
The historical recommendation interest points refer to interest points which are recommended to various objects to be recommended historically. The voting information refers to selection information of the object to be recommended for the historical recommendation interest points, for example, three interest points are recommended to the object to be recommended by the history, then one of the interest points is selected by each object to be recommended as a final interest point, and the selection information of each object to be recommended after each recommendation is recorded, so that the voting information is obtained.
Specifically, the server obtains voting information of each object to be recommended for historical recommendation interest points recorded in the database, calculates recommendation weights corresponding to the objects to be recommended according to the voting information, for example, 10 historical records exist, and counts the number of times that the interest point selected by each object to be recommended is the final interest point according to whether the interest point selected by the object to be recommended is consistent with the final interest point, for example, the number of times that the interest point selected by the object to be recommended is the final interest point is 5 times, and the probability of successful selection is one half. And calculating to obtain the probability of successful selection of each object to be recommended, and then normalizing the probability of successful selection to a value between 0 and 1 to obtain the recommendation weight corresponding to each object to be recommended. In this embodiment, the recommendation weight is calculated by the voting information, and the obtained recommendation weight can be more accurate.
In an embodiment, as shown in fig. 10, in step 908, sorting the average candidate recommendation degrees corresponding to the candidate interest points, and selecting a preset number of candidate interest points as target recommendation interest points according to the sorting result, includes:
step 1002, the average candidate recommendation degrees corresponding to the candidate interest points are ranked to obtain a first ranking result.
And 1004, respectively calculating the sum of the distances from each candidate interest point to each initial position, and sorting each candidate interest point according to the sum of the distances to obtain a second sorting result.
The first ranking result is obtained by ranking the candidate interest points according to the average candidate recommendation degree. The sum of distances refers to the sum of the distances of each candidate point of interest to all starting positions. The second ranking result is a result obtained by ranking the candidate interest points according to the sum of the distances.
Specifically, the server sorts the candidate interest points in sequence from large to small according to the average candidate recommendation degree to obtain a first sorting result. Then the server calculates the sum of the distances from each candidate interest point to each starting position respectively, namely calculates the distance from each candidate interest point to each starting position, and then calculates the sum of all the distances to obtain the sum of the distances. And then sequencing the candidate interest points in sequence according to the distance sum from small to large to obtain a second sequencing result.
Step 1006, determining a final recommended interest point from the candidate interest points based on the first ranking result and the second ranking result.
Specifically, the server may select the candidate interest points ranked in the first ranking result and ranked in the second ranking result at the same time as the final recommended interest points. For example, the server searches whether there is the same candidate interest point in the first three candidate interest points of the first sorting result and the second sorting result, when there is the same candidate interest point, the same candidate interest point is used as the final recommended interest point, when there is no same candidate interest point, whether there is the same candidate interest point is searched in the first five candidate interest points, and the cyclic search is continuously performed until the same candidate interest point is found, and the searched same candidate interest point is used as the final recommended interest point. The final recommendation interest point refers to an interest point to be finally recommended, and there may be a plurality of interest points.
In the embodiment, the first ranking result is obtained according to the average candidate recommendation degree, the second ranking result is obtained according to the distance sum, and then the final recommended interest point is determined from the candidate interest points according to the first ranking result and the second ranking result, so that the accuracy of the obtained final recommended interest point is improved.
In one embodiment, as shown in fig. 11, step 212, determining a target recommendation interest point based on the interest point attribute information and the recommendation degree of each to-be-recommended interest point, includes:
step 1102, determining candidate interest points from the interest points to be recommended based on the interest point attribute information corresponding to the objects to be recommended, and searching for candidate recommendation degrees corresponding to the candidate interest points from the recommendation degrees of the interest points to be recommended.
And 1104, acquiring voting information of each object to be recommended for historical recommendation interest points, and extracting voting characteristics based on the voting information.
The voting characteristics are used for representing voting information corresponding to each object to be recommended. The voting information refers to voting information of each object to be recommended for historical recommendation interest points, namely, the interest points of the final party are determined from the historical recommendation interest points by each object to be recommended in a voting mode.
Specifically, the server may acquire, in the database, the voting information of each object to be recommended to the historical recommendation interest point each time. And then, extracting features of the voting information to obtain voting features. Namely, the recommendation weight corresponding to each object to be recommended can be calculated according to the voting information, and then the recommendation weight is directly used as the voting characteristic.
And step 1106, calculating the sum of the distances from each candidate interest point to each starting position, and obtaining distance features based on the sum of the distances.
Wherein the distance features are used for characterizing the sum of distances corresponding to each candidate interest point.
Specifically, the server directly uses the distance sum of each candidate interest point to each starting position as the distance feature.
Step 1108, obtaining the number of objects to be recommended, and calculating the sum of each candidate recommendation degree corresponding to each candidate interest point to obtain the sum of the candidate recommendation degrees corresponding to each candidate interest point.
Step 1110, performing average calculation based on the sum of the candidate recommendation degrees and the number of the objects to be recommended to obtain an average candidate recommendation degree corresponding to each candidate interest point, and obtaining recommendation degree characteristics based on the average candidate recommendation degree.
The number of objects to be recommended refers to the total number of each object to be recommended, and may be, for example, the total number of users in a party. The sum of the candidate recommendation degrees refers to the sum of all candidate recommendation degrees corresponding to the candidate interest points. The recommendation degree feature is used for representing the average candidate recommendation degree of each candidate interest point.
Specifically, the server may count the number of objects to be recommended obtained by each object to be recommended. Then, calculating the sum of the candidate recommendation degrees corresponding to the objects to be recommended corresponding to each candidate interest point to obtain the candidate recommendation degree sum corresponding to each candidate interest point, then, calculating the ratio of the candidate recommendation degree sum to the number of the objects to be recommended by the server, namely, performing average calculation to obtain the average candidate recommendation degree corresponding to each candidate interest point, and then, taking the average candidate recommendation degree corresponding to each candidate interest point as recommendation degree characteristics.
Step 1112, inputting the voting features, the distance features and the recommendation degree features into a first interest point recommendation model to obtain target candidate recommendation degrees corresponding to each output candidate interest point, and determining model recommendation interest points from each candidate interest point based on the target candidate recommendation degrees, wherein the first interest point recommendation model is obtained by training through a machine learning algorithm by using training data, and the training data comprises a training initial position, training interest point attribute information and training recommendation interest point labels.
The training starting position refers to a starting position corresponding to each object to be recommended for training used in training the first interest point recommendation model. The object to be recommended for training refers to an object to be recommended during training. The training interest point attribute information refers to interest point attribute information corresponding to an object to be recommended during training. The training recommended interest point label is a label used in training to determine whether a historical interest point is recommended or not, and comprises a recommended label and an unrecommended label. The recommendation tag is used for indicating that the corresponding historical interest point is historically recommended. The non-recommended tags are used to indicate that the corresponding historical points of interest are historically non-recommended. The model recommendation interest point refers to an interest point obtained by recommending the interest point by using the first interest point recommendation model.
Specifically, the server performs training in advance by using training data based on a Machine learning algorithm to obtain a first interest point recommendation model, where the Machine learning algorithm may be an SVM (Support Vector Machine) algorithm, a neural network algorithm, or the like. The server can acquire training data, and then can extract training distance features, training voting features and training average candidate recommendation degrees by using the training data, wherein the server can calculate a training starting range by using each training starting position, then determines training candidate recommendation interest points from the training starting range according to the training interest point attribute information, then calculates the training distance sum from each training candidate recommendation interest point to each training starting position, and takes the training distance sum corresponding to each training candidate recommendation interest point as the training distance features. The server can also acquire the voting information of the training object to be recommended for the historical recommendation interest points, and extract the training voting characteristics based on the voting information. The server can also obtain the number of the objects to be recommended, and calculate the sum of the recommendation degrees of the training candidates corresponding to each training candidate interest point to obtain the sum of the recommendation degrees of the training candidates corresponding to each training candidate interest point. And performing average calculation based on the sum of the training candidate recommendation degrees and the number of the objects to be recommended to be trained to obtain the training average candidate recommendation degrees corresponding to the training candidate interest points, and obtaining the training recommendation degree characteristic based on the training average candidate recommendation degrees. And then the server takes the training distance characteristic, the training voting characteristic and the training average candidate recommendation degree as the input of a model established by a machine learning algorithm, takes the training recommendation interest point labels as labels for training, and obtains a first interest point recommendation model when the training is finished. The server then deploys and uses the first point of interest recommendation model.
When the target candidate recommendation degree corresponding to each candidate interest point is output, the server splices the voting characteristics, the distance characteristics and the recommendation degree characteristics to obtain spliced characteristics, and inputs the spliced characteristics into the first interest point recommendation model to obtain the target candidate recommendation degree corresponding to each output candidate interest point. Then the server can sort the candidate interest points from large to small according to the target candidate recommendation degree, and then select a preset number of candidate interest points as model recommendation interest points according to the sorting result
In the embodiment, the accuracy of obtaining the target candidate recommendation degrees corresponding to the candidate interest points is improved by obtaining the voting characteristics, the distance characteristics and the recommendation degree characteristics and then calculating the voting characteristics, the distance characteristics and the recommendation degree characteristics by using the first interest point recommendation model, and then the server selects the interest points to be recommended from the candidate interest points according to the target candidate recommendation degrees, so that the selected interest points to be recommended can be more accurate.
In one embodiment, the method for recommending points of interest further comprises:
inputting each candidate interest point into a second interest point recommendation model to obtain an output target recommendation interest point, wherein the second interest point recommendation model is obtained by training history candidate interest points and corresponding history recommendation interest points based on a deep neural network algorithm, and the history candidate interest points are obtained by calculating history initial positions and history interest point attribute information corresponding to each history object to be recommended.
The historical recommendation interest point food refers to an interest point which is recommended to each historical object to be recommended in a historical mode. The history initial position is the initial position corresponding to each history object to be recommended when the point of interest recommendation is performed on each history object to be recommended. The historical interest point attribute information refers to interest point attribute information corresponding to each historical object to be recommended.
Specifically, the server may also use a deep neural network algorithm based on history candidate interest points and corresponding history recommended interest points to train to obtain a second interest point recommendation model. That is, the server may obtain a history initial position and history interest point attribute information corresponding to each history object to be recommended, then calculate and obtain a history initial range according to the history initial position, then screen and obtain history candidate interest points from each history interest point in the history initial range according to the history interest point attribute information, and then use the history candidate interest points as an input of a model established by a Deep Neural Network algorithm, where the Deep Neural Network algorithm may be a DCN (Deep & Cross Network) algorithm, a CNN (Convolutional Neural Network, probabilistic Neural Networks) algorithm, an RNN (Recurrent Neural Network, RNN) algorithm, or the like. And then training the historical recommended interest points as labels of the established model until the training is finished to obtain a second interest point recommended model, wherein the training is finished when the training finishing condition is met. The training completion condition includes at least one of the loss function value reaching a preset threshold, the maximum iteration number reaching, and no change of the model parameter. In one embodiment, the second point of interest recommendation model may be the above-mentioned each
And deploying and using the trained second interest point recommendation model by the server, acquiring each candidate interest point by the server when the server is used, and inputting each candidate interest point into the second interest point recommendation model for calculation to obtain an output target recommendation interest point.
In the embodiment, the server uses the pre-trained second interest point recommendation model to recommend the interest points, so that the efficiency and the accuracy of obtaining the target recommended interest points are improved.
In an embodiment, as shown in fig. 12, a method for showing a point of interest is provided, which is described by taking the method as an example for being applied to the terminal in fig. 1, and includes the following steps:
step 1202, responding to an interest point recommending event triggered by a target object to be recommended, and displaying an interest point recommending interface, wherein an initial position corresponding to each object to be recommended is displayed in the interest point recommending interface, and each object to be recommended comprises the target object to be recommended.
The interest point recommending event refers to an event triggered by a terminal for interest point recommendation, the target object to be recommended refers to an object needing interest point recommendation, and the target object to be recommended is one of the objects to be recommended. The point of interest recommendation interface refers to an interface for performing point of interest recommendation.
Specifically, the terminal receives an interest point recommendation event triggered by a target object to be recommended, wherein the target object to be recommended can start an application program recommended by a meeting place in the terminal, and the terminal receives the interest point recommendation event triggered by the target object to be recommended by performing an interest point recommendation operation in the application program, for example, clicking an interest point recommendation button. At the moment, the terminal responds to an interest point recommending event triggered by the target object to be recommended and displays an interest point recommending interface, wherein the interest point recommending interface displays an initial position corresponding to each object to be recommended, and each object to be recommended comprises the target object to be recommended. The terminal can acquire the starting position corresponding to each object to be recommended from the server and then display the starting position in the interest point recommendation interface.
In one embodiment, the terminal may obtain, from the server, the start positions corresponding to the objects to be recommended except for the target object to be recommended, and display, in the interest point recommendation interface, the start positions corresponding to the objects to be recommended except for the target object to be recommended. And then the terminal acquires the initial position of the target object to be recommended and displays the initial position of the target object to be recommended in the interest point recommendation interface. The terminal can acquire the initial position of the target object to be recommended input through the interest point recommendation interface, or the terminal can acquire the position of the target object to be recommended as the initial position through a positioning technology.
In a specific embodiment, as shown in fig. 13, a schematic diagram of a point of interest recommendation interface is shown, in which starting positions of three objects to be recommended are shown, including a starting position 1302, a starting position 1304, and a starting position 1306.
Step 1204, responding to the interest point attribute information selection event triggered by the interest point recommendation interface, and displaying an interest point attribute information acquisition interface.
The interest point attribute information selection event refers to an event for selecting the interest point attribute information. The interest point attribute information acquisition interface is used for acquiring interest point attribute information corresponding to the target object to be recommended.
Specifically, the terminal receives an interest point attribute information selection operation of a target object to be recommended through an interest point recommendation interface, and triggers an interest point attribute information selection event. The point of interest attribute information selection operation refers to an operation for performing point of interest attribute information selection, and includes but is not limited to a click operation, a press operation, a slide operation, and the like. At this time, the terminal responds to an interest point attribute information selection event triggered through the interest point recommendation interface, an interest point attribute information acquisition interface is displayed, and the target object to be recommended can select the interest point attribute information through the interest point attribute information acquisition interface and also can acquire the interest point attribute information of acquaintances of the interface through the interest point attribute information.
In a specific embodiment, as shown in fig. 14, for an interest point attribute information obtaining interface diagram, a user may select interest point attribute information through the interest point attribute information obtaining interface, for example, may select price information of a restaurant, a xiangcai, a slightly spicy dish, less than 50, and the like.
In step 1206, in response to the point of interest attribute information confirmation event triggered by the point of interest attribute information acquisition interface, target point of interest attribute information corresponding to the target object to be recommended is acquired, and the target point of interest attribute information is sent to the server.
The interest point attribute information confirmation event is used for confirming the interest point attribute information. The target interest point attribute information refers to interest point attribute information corresponding to a target object to be recommended.
Specifically, the terminal receives an interest point attribute information confirmation operation of a target object to be recommended and a triggered interest point attribute information confirmation event through an interest point attribute information acquisition interface, wherein the confirmation operation includes but is not limited to a click operation, a press operation, a slide operation, a drag operation and the like. And then the terminal responds to an interest point attribute information confirmation event triggered by the interest point attribute information acquisition interface, acquires target interest point attribute information corresponding to the target object to be recommended, and then sends the target interest point attribute information to the server. In an embodiment, the terminal may also send the target interest point attribute information and the start position corresponding to the target object to be recommended to the server at the same time. In the specific embodiment of fig. 14, the user may confirm the event by clicking the confirmation button, and the terminal acquires the target interest point attribute information by responding to the interest point attribute information confirmation event, and sends the target interest point attribute information to the server.
And step 1208, displaying the target recommended interest points returned by the server, wherein the target recommended interest points are obtained by determining candidate ranges of the interest points to be recommended by the server according to the initial positions corresponding to the objects to be recommended, determining the candidate interest points from the candidate ranges of the interest points to be recommended based on the attribute information of the interest points corresponding to the objects to be recommended, and performing matrix decomposition prediction on the candidate recommendation degrees corresponding to the candidate interest points based on the candidate recommendation degrees corresponding to the candidate interest points, wherein the candidate recommendation degrees corresponding to the candidate interest points are searched from the target interest point recommendation degree matrix which is obtained by using the original interest point recommendation degree matrix.
Specifically, when receiving the attribute information of the point of interest sent by each object to be recommended, the server obtains the starting position corresponding to each object to be recommended. Then, determining a candidate range of interest points to be recommended according to the initial position, screening each candidate interest point from the candidate range of interest points to be recommended according to the attribute information of the interest point sent by each object to be recommended, and then searching candidate recommendation degrees corresponding to each candidate interest point from a target interest point recommendation degree matrix, wherein the target interest point recommendation degree matrix is obtained by performing matrix decomposition prediction by using an original interest point recommendation degree matrix, and the original interest point recommendation degree matrix is obtained by performing matrix establishment according to each interest point to be recommended in the candidate range of interest points to be recommended and the historical recommendation degrees of historical interest points corresponding to each object to be recommended. At this time, the server selects a preset number of candidate interest points as each target recommendation interest point from large to small in sequence according to the candidate recommendation degree corresponding to each candidate interest point, and then sends each target recommendation interest point to the terminal of each object to be recommended. And when the terminal of the target object to be recommended receives each target recommendation interest point sent by the server, displaying each target recommendation interest point. The terminal can display each target recommendation interest point in the interest point recommendation interface, can display each target recommendation interest point through a display list, can display through a target recommendation interest point display interface, and the like.
In one embodiment, when the terminal acquires each target recommendation interest point, only one target recommendation interest point may be displayed, and when the target recommendation interest point switching operation of the user is received, another target recommendation interest point is displayed.
In the method for displaying the interest points, the terminal displays an interest point recommending interface by responding to an interest point recommending event triggered by a target object to be recommended, responds to an interest point attribute information selecting event triggered by the interest point recommending interface, displays an interest point attribute information acquiring interface, responds to an interest point attribute information confirming event triggered by the interest point attribute information acquiring interface, acquires target interest point attribute information corresponding to the target object to be recommended, sends the target interest point attribute information to the server, and displays the target recommended interest points returned by the server, so that the target object to be recommended can acquire accurate target recommended interest points, the target recommended interest points displayed by the terminal are more accurate, the terminal can display the initial positions of all the objects to be recommended, namely, common content can be synchronously displayed, and then the terminal can acquire the interest point attribute information corresponding to the target object to be recommended through the interest point attribute information acquiring interface The sex information, namely the terminal can also show the personalized content, thereby being convenient for the use of the user.
In a specific embodiment, as shown in fig. 15, a method for recommending a point of interest is provided, which specifically includes the following steps:
step 1502, obtaining the initial position and the interest point attribute information corresponding to each object to be recommended, screening from each initial position to obtain each reference initial position, and determining an initial range based on each reference initial position.
Step 1504, determining each reference range based on each reference starting position and the preset circle radius, and combining each reference range and the starting range to obtain the candidate range of the interest point to be recommended.
Step 1506, obtaining a historical recommendation degree of the historical interest point corresponding to each object to be recommended, obtaining a matrix column attribute based on each interest point to be recommended and the historical interest point, obtaining a matrix row attribute based on each object to be recommended, taking the historical recommendation degree of the historical interest point as a matrix value, and obtaining an original interest point recommendation degree matrix based on the matrix column attribute, the matrix row attribute and the matrix value.
Step 1508, performing matrix decomposition on the original interest point recommendation degree matrix according to a preset decomposition relation to obtain a first decomposition matrix and a second decomposition matrix, calculating a product of the first decomposition matrix and the second decomposition matrix to obtain a reconstructed interest point recommendation degree matrix, and calculating a sum of squares of errors of each recommendation degree in the original interest point recommendation degree matrix and a corresponding reconstructed recommendation degree in the reconstructed interest point recommendation degree matrix.
Step 1510, calculating first penalty item information corresponding to the first decomposition matrix based on a preset first penalty coefficient, calculating second penalty item information corresponding to the second decomposition matrix based on a preset second penalty coefficient, obtaining loss information based on the error square sum, the first penalty item information and the second penalty item information, and updating the first decomposition matrix and the second decomposition matrix based on the loss information.
Step 1512, the step of calculating the product of the first decomposition matrix and the second decomposition matrix is returned, and the step of reconstructing the interest point recommendation degree matrix is iteratively executed until the target first decomposition matrix and the target second decomposition matrix are obtained when the loss information meets the preset loss condition, and the product of the target first decomposition matrix and the target second decomposition matrix is calculated, so that the target interest point recommendation degree matrix is obtained.
Step 1514, determining candidate interest points from the interest points to be recommended based on the interest point attribute information corresponding to the objects to be recommended, and searching for candidate recommendation degrees corresponding to the candidate interest points from the recommendation degrees of the interest points to be recommended;
step 1516, obtaining the number of objects to be recommended, calculating the sum of each candidate recommendation degree corresponding to each candidate interest point, obtaining the sum of the candidate recommendation degrees corresponding to each candidate interest point, performing average calculation based on the sum of the candidate recommendation degrees and the number of objects to be recommended, obtaining the average candidate recommendation degree corresponding to each candidate interest point, sorting the average candidate recommendation degrees corresponding to each candidate interest point, and selecting a preset number of candidate interest points as target recommendation interest points according to the sorting result.
The application also provides an application scenario, and the interest point recommendation method is applied to the application scenario. Specifically, the application of the interest point recommendation method in the application scenario is as follows:
in an application having a function of recommending a meeting place, when a plurality of users use the application to recommend a meeting place, a flow of recommending a meeting place for a server is shown in fig. 16. Specifically, the method comprises the following steps: each user submits meeting place attribute information and departure position information to the server through the application program. The server obtains the meeting place attribute information and the departure position information of a plurality of users, performs distance screening, namely calculates a convex hull by using the departure position information of the plurality of users based on a divide-and-conquer algorithm, and obtains a vertex on the convex hull. Then, area determination is carried out by taking a vertex on the convex hull as a center and taking the radius as a preset distance, such as 20 meters, so as to obtain a circular area range, then an initial range is obtained according to the area range and the circular area range of the convex hull, and the server determines each candidate meeting place from the initial range according to the attribute information of the meeting places of a plurality of users. Then, the server obtains a target gathering place scoring matrix through multi-element recommendation calculation based on matrix decomposition, namely historical evaluation scores of historical gathering places corresponding to all users are obtained, matrix building is conducted based on historical evaluation scores of all to-be-recommended interest points and historical gathering places corresponding to all to-be-recommended objects, and an original interest point scoring matrix is obtained. And performing matrix decomposition prediction based on the original interest point recommendation degree matrix to obtain a target meeting place score matrix, wherein the target meeting place score matrix comprises the prediction scores of all users on all meeting places to be recommended. And then the server searches candidate scores corresponding to the candidate party positions from the recommendation degrees corresponding to the to-be-recommended party positions of the target party position score matrix. And then the server determines a target recommended gathering place from the candidate gathering places based on the size of the candidate scores, and sends the target recommended gathering place to the terminals corresponding to the users.
It should be understood that although the various steps in the flowcharts of fig. 2, 3, 5, 7-12, 15 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least some of the steps in fig. 2, 3, 5, 7-12, 15 may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, which are not necessarily performed in sequence, but may be performed alternately or alternately with other steps or at least some of the other steps.
In one embodiment, as shown in fig. 17, an apparatus 1700 for point of interest recommendation is provided, which may be a part of a computer device using a software module or a hardware module, or a combination of the two modules, and specifically includes: an initial range determining module 1702, a candidate range obtaining module 1704, an original matrix obtaining module 1706, a matrix decomposition module 1708, and an interest point recommending module 1710, wherein:
an initial range determining module 1702, configured to obtain initial positions and interest point attribute information corresponding to each object to be recommended, filter the initial positions to obtain reference initial positions, and determine an initial range based on the reference initial positions;
a candidate range obtaining module 1704, configured to determine a corresponding reference range based on each reference starting position, and obtain a candidate range of interest points to be recommended according to the reference range and the starting range, where the candidate range of interest points to be recommended includes each interest point to be recommended;
an original matrix obtaining module 1706, configured to obtain an original interest point recommendation degree matrix corresponding to each interest point to be recommended;
a matrix decomposition module 1708, configured to perform matrix decomposition prediction based on the original interest point recommendation degree matrix to obtain a target interest point recommendation degree matrix, where the target interest point recommendation degree matrix includes recommendation degrees of each object to be recommended to each interest point to be recommended;
the interest point recommending module 1712 is configured to determine a target recommending interest point based on the interest point attribute information corresponding to each object to be recommended and the recommendation degree of each interest point to be recommended, and send the target recommending interest point to a terminal corresponding to each object to be recommended.
In one embodiment, the starting range determination module 1702 includes:
a reference position determination unit for determining a first reference start position and a second reference start position based on the position coordinates corresponding to the respective start positions;
the dividing unit is used for determining a first straight line based on the first reference initial position and the second reference initial position, and dividing each initial position based on the first straight line to obtain a first part initial position and a second part initial position;
the screening unit is used for screening from the initial positions of the first parts to obtain reference initial positions of the first parts and screening from the initial positions of the second parts to obtain reference initial positions of the second parts;
a position obtaining unit, configured to obtain each reference starting position based on the first reference starting position, the second reference starting position, each first partial reference starting position, and each second partial reference starting position, and determine the starting range based on each reference starting position.
In one embodiment, the screening unit is further configured to calculate a vertical distance between each of the first part start positions and the first straight line, and determine a third reference start position from the first part start positions based on the vertical distance; determining a second straight line based on a third reference starting position and the first reference starting position, dividing the first part starting position based on the second straight line to obtain a third part starting position, determining a third straight line based on the third reference starting position and the second reference starting position, and dividing the first part starting position based on the third straight line to obtain a fourth part starting position; and respectively taking the starting position of the third part and the starting position of the fourth part as the starting position of the first part, respectively taking the second straight line and the third straight line as the first straight line, returning to calculate the vertical distance between each starting position in the starting positions of the first part and the first straight line, and determining a third reference starting position from the starting positions of the first part based on the vertical distance.
In one embodiment, the candidate range derivation module 1704 is further configured to determine each reference range based on each reference starting position and a preset circle radius; and combining each reference range and the initial range to obtain a candidate range of the interest point to be recommended.
In an embodiment, the original matrix obtaining module 1706 is further configured to obtain a matrix column attribute based on each to-be-recommended interest point and the historical interest points, and obtain a matrix row attribute based on each to-be-recommended object; and taking the historical recommendation degree of the historical interest point as a matrix value, and obtaining an original interest point recommendation degree matrix based on the matrix column attribute, the matrix row attribute and the matrix value.
In one embodiment, the matrix decomposition module 1708 includes:
the analysis unit is used for performing matrix analysis on the original interest point recommendation degree matrix according to a preset decomposition relation to obtain a first decomposition matrix and a second decomposition matrix, wherein the first decomposition matrix is used for representing the recommendation degree relation between each object to be recommended and each attribute in the interest point attribute information, and the second decomposition matrix is used for representing the weight relation between each attribute in the interest point attribute information and the interest point;
the reconstruction unit is used for calculating the product of the first decomposition matrix and the second decomposition matrix to obtain a reconstructed interest point recommendation degree matrix;
the loss calculation unit is used for calculating loss information of the original interest point recommendation degree matrix and the reconstructed interest point recommendation degree matrix and updating the first decomposition matrix and the second decomposition matrix based on the loss information;
the iteration unit is used for calculating the product of the first decomposition matrix and the second decomposition matrix in a return mode to obtain a step of reconstructing the recommendation degree matrix of the interest point, and performing iteration until loss information meets a preset loss condition to obtain a target first decomposition matrix and a target second decomposition matrix;
and the target matrix obtaining unit is used for calculating the product of the target first decomposition matrix and the target second decomposition matrix to obtain a target interest point recommendation degree matrix.
In one embodiment, the loss calculating unit is further configured to calculate a sum of squares of errors of each recommendation degree in the original interest point recommendation degree matrix and a corresponding reconstruction recommendation degree in the reconstruction interest point recommendation degree matrix; calculating first punishment item information corresponding to the first decomposition matrix based on a preset first punishment coefficient, and calculating second punishment item information corresponding to the second decomposition matrix based on a preset second punishment coefficient; and obtaining loss information based on the error square sum, the first penalty item information and the second penalty item information.
In one embodiment, the interest point recommending module 1712 is further configured to obtain the number of objects to be recommended, and calculate the sum of each candidate recommendation degree corresponding to each candidate interest point to obtain the sum of the candidate recommendation degrees corresponding to each candidate interest point; performing average calculation based on the sum of the candidate recommendation degrees and the number of objects to be recommended to obtain the average candidate recommendation degree corresponding to each candidate interest point; and sorting the average candidate recommendation degrees corresponding to the candidate interest points, and selecting a preset number of candidate interest points as target recommendation interest points according to a sorting result.
In one embodiment, the interest point recommending module 1712 is further configured to obtain a recommendation weight corresponding to each object to be recommended, and perform weighted calculation on the candidate recommendation degrees corresponding to each object to be recommended according to the recommendation weight to obtain each weighted candidate recommendation degree; and calculating the sum of the weighted candidate recommendation degrees of each candidate interest point to obtain the candidate recommendation degree sum of each candidate interest point.
In an embodiment, the interest point recommending module 1712 is further configured to obtain voting information of each object to be recommended for the historical recommendation interest point, and calculate a recommendation weight corresponding to each object to be recommended based on the voting information.
In one embodiment, the interest point recommending module 1712 is further configured to rank the average candidate recommendation degrees corresponding to the candidate interest points to obtain a first ranking result; respectively calculating the distance sum of each candidate interest point to each initial position, and sequencing each candidate interest point according to the distance sum to obtain a second sequencing result; and determining a final recommended interest point from the candidate interest points based on the first sorting result and the second sorting result.
In one embodiment, the interest point recommending module 1712 is further configured to obtain voting information of each object to be recommended for historical recommendation interest points, and extract voting features based on the voting information; calculating the sum of the distances from each candidate interest point to each initial position, and obtaining distance features based on the sum of the distances; acquiring the number of objects to be recommended, and calculating the sum of each candidate recommendation degree corresponding to each candidate interest point to obtain the sum of the candidate recommendation degrees corresponding to each candidate interest point; performing average calculation based on the sum of the candidate recommendation degrees and the number of objects to be recommended to obtain an average candidate recommendation degree corresponding to each candidate interest point, and obtaining recommendation degree characteristics based on the average candidate recommendation degree; inputting the voting characteristics, the distance characteristics and the recommendation degree characteristics into a first interest point recommendation model to obtain target candidate recommendation degrees corresponding to each output candidate interest point, determining model recommendation interest points from each candidate interest point based on the target candidate recommendation degrees, wherein the first interest point recommendation model is obtained by training through training data based on a machine learning algorithm, and the training data comprises a training initial position, training interest point attribute information and training recommendation interest point labels.
In one embodiment, the point of interest recommending apparatus 1700 further includes:
and the model recommendation module is used for inputting each candidate interest point into a second interest point recommendation model to obtain an output target recommendation interest point, the second interest point recommendation model is obtained by training through historical candidate interest points and corresponding historical recommendation interest points based on a deep neural network algorithm, and the historical candidate interest points are obtained by calculating through historical initial positions and historical interest point attribute information corresponding to each historical object to be recommended.
In one embodiment, as shown in fig. 18, there is provided a point of interest exhibition apparatus 1800, which may be a part of a computer device using software modules or hardware modules, or a combination of both, and specifically includes: a recommendation interface display module 1802, an information acquisition interface display module 1804, an information sending module 1806, and an interest point display module 1808, wherein:
a recommendation interface display module 1802, configured to respond to an interest point recommendation event triggered by a target object to be recommended, and display an interest point recommendation interface, where an initial position corresponding to each object to be recommended is displayed in the interest point recommendation interface, and each object to be recommended includes the target object to be recommended;
the information obtaining interface display module 1804 is configured to respond to the point of interest attribute information selection event triggered by the point of interest recommendation interface, and display the point of interest attribute information obtaining interface;
an information sending module 1806, configured to respond to an interest point attribute information confirmation event triggered through the interest point attribute information acquisition interface, acquire target interest point attribute information corresponding to a target object to be recommended, and send the target interest point attribute information to a server;
an interest point display module 1808, configured to display a target recommended interest point returned by the server, where the target recommended interest point is obtained by determining, by the server, a candidate range of interest points to be recommended according to an initial position corresponding to each object to be recommended, determining, based on interest point attribute information corresponding to each object to be recommended, each candidate interest point from the candidate range of interest points to be recommended, and based on a candidate recommendation degree corresponding to each candidate interest point, where the candidate recommendation degree corresponding to each candidate interest point is found from a target interest point recommendation degree matrix, and the target interest point recommendation degree matrix is obtained by performing matrix decomposition prediction using an original interest point recommendation degree matrix.
For the specific limitations of the point of interest recommendation device and the point of interest display device, see the above limitations on the point of interest recommendation method, which are not described herein again. All or part of the modules in the interest point recommending device and the interest point showing device can be realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in fig. 19. The computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for storing historical recommendation data, a target interest point recommendation degree matrix, each starting position and interest point attribute information and the like. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a point of interest recommendation method.
In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 20. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless communication can be realized through WIFI, an operator network, NFC (near field communication) or other technologies. The computer program is executed by a processor to implement a point of interest exposure method. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.
It will be appreciated by those skilled in the art that the configurations shown in fig. 19 and 20 are block diagrams of only some of the configurations relevant to the present application, and do not constitute a limitation on the computing devices to which the present application may be applied, and that a particular computing device may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is further provided, which includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the above method embodiments when executing the computer program.
In an embodiment, a computer-readable storage medium is provided, in which a computer program is stored which, when being executed by a processor, carries out the steps of the above-mentioned method embodiments.
In one embodiment, a computer program product or computer program is provided that includes computer instructions stored in a computer-readable storage medium. The computer instructions are read by a processor of a computer device from a computer-readable storage medium, and the computer instructions are executed by the processor to cause the computer device to perform the steps in the above-mentioned method embodiments.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database or other medium used in the embodiments provided herein can include at least one of non-volatile and volatile memory. Non-volatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical storage, or the like. Volatile Memory can include Random Access Memory (RAM) or external cache Memory. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (30)
1. A method for point of interest recommendation, the method comprising:
acquiring initial positions and interest point attribute information corresponding to objects to be recommended, screening the initial positions to obtain reference initial positions, and determining an initial range based on the reference initial positions;
determining a corresponding reference range based on each reference starting position, and obtaining a candidate range of interest points to be recommended according to the reference range and the starting range, wherein the candidate range of interest points to be recommended comprises each interest point to be recommended;
acquiring an original interest point recommendation degree matrix corresponding to each interest point to be recommended, wherein the original interest point recommendation degree matrix is established by using the interest points to be recommended, historical interest points and historical recommendation degrees of each object to be recommended corresponding to the historical interest points;
performing matrix decomposition prediction based on the original interest point recommendation degree matrix to obtain a target interest point recommendation degree matrix, wherein the target interest point recommendation degree matrix comprises recommendation degrees of the objects to be recommended to the interest points to be recommended;
and determining a target recommendation interest point based on the interest point attribute information corresponding to each object to be recommended and the recommendation degree of each interest point to be recommended, and sending the target recommendation interest point to a terminal corresponding to each object to be recommended.
2. The method of claim 1, wherein the screening from the respective starting positions to obtain respective reference starting positions, and determining a starting range based on the respective reference starting positions comprises:
determining a first reference starting position and a second reference starting position based on the position coordinates corresponding to the starting positions;
determining a first straight line based on the first reference starting position and the second reference starting position, and dividing each starting position based on the first straight line to obtain a first part starting position and a second part starting position;
screening from the initial positions of the first parts to obtain reference initial positions of the first parts, and screening from the initial positions of the second parts to obtain reference initial positions of the second parts;
and obtaining each reference starting position based on the first reference starting position, the second reference starting position, each first partial reference starting position and each second partial reference starting position, and determining a starting range based on each reference starting position.
3. The method of claim 2, wherein said filtering from the first portion starting positions to obtain respective first portion reference starting positions comprises:
calculating a vertical distance from each of the first partial start positions to the first straight line, and determining a third reference start position from the first partial start positions based on the vertical distance;
determining a second straight line based on the third reference starting position and the first reference starting position, dividing the first part starting position based on the second straight line to obtain a third part starting position, determining a third straight line based on the third reference starting position and the second reference starting position, and dividing the first part starting position based on the third straight line to obtain a fourth part starting position;
and respectively taking the third part starting position and the fourth part starting position as first part starting positions, respectively taking the second straight line and the third straight line as first straight lines, returning and calculating the vertical distance between each starting position in the first part starting positions and the first straight lines, and executing the step of determining a third reference starting position from the first part starting positions based on the vertical distance until the first part starting positions do not contain the starting positions to obtain each first part reference starting position.
4. The method according to claim 1, wherein the determining a corresponding reference range based on each reference starting position and obtaining a candidate range of interest points to be recommended according to the reference range and the starting range comprises:
determining each reference range based on each reference starting position and a preset circle radius;
and combining the reference ranges and the initial range to obtain the candidate range of the interest point to be recommended.
5. The method according to claim 1, wherein the obtaining of the original interest point recommendation degree matrix corresponding to each interest point to be recommended includes:
acquiring historical recommendation degrees of historical interest points corresponding to the objects to be recommended, obtaining matrix column attributes based on the interest points to be recommended and the historical interest points, and obtaining matrix row attributes based on the objects to be recommended;
and taking the historical recommendation degree of the historical interest points as a matrix value, and obtaining the original interest point recommendation degree matrix based on the matrix column attribute, the matrix row attribute and the matrix value.
6. The method of claim 1, wherein performing matrix decomposition prediction based on the original interest point recommendation degree matrix to obtain a target interest point recommendation degree matrix comprises:
performing matrix decomposition on the original interest point recommendation degree matrix according to a preset decomposition relationship to obtain a first decomposition matrix and a second decomposition matrix, wherein the first decomposition matrix is used for representing the recommendation degree relationship between each object to be recommended and each attribute in the interest point attribute information, and the second decomposition matrix is used for representing the weight relationship between each attribute in the interest point attribute information and the interest point;
calculating the product of the first decomposition matrix and the second decomposition matrix to obtain a reconstructed interest point recommendation degree matrix;
calculating loss information of the original interest point recommendation degree matrix and the reconstructed interest point recommendation degree matrix, and updating the first decomposition matrix and the second decomposition matrix based on the loss information;
the step of calculating the product of the first decomposition matrix and the second decomposition matrix is returned to obtain a reconstructed interest point recommendation degree matrix, and the step is executed iteratively until the target first decomposition matrix and the target second decomposition matrix are obtained when the loss information meets the preset loss condition;
and calculating the product of the target first decomposition matrix and the target second decomposition matrix to obtain the target interest point recommendation degree matrix.
7. The method of claim 6, wherein the calculating the loss information of the original point of interest recommendation matrix and the reconstructed point of interest recommendation matrix comprises:
calculating the sum of the squares of errors of all recommendation degrees in the original interest point recommendation degree matrix and corresponding reconstruction recommendation degrees in the reconstruction interest point recommendation degree matrix;
calculating first penalty item information corresponding to the first decomposition matrix based on a preset first penalty coefficient, and calculating second penalty item information corresponding to the second decomposition matrix based on a preset second penalty coefficient;
and obtaining the loss information based on the error square sum, the first penalty item information and the second penalty item information.
8. The method according to claim 1, wherein the determining a target recommendation interest point based on the interest point attribute information and the recommendation degree of each interest point to be recommended comprises:
determining candidate interest points from the interest points to be recommended based on the interest point attribute information corresponding to the objects to be recommended, and searching candidate recommendation degrees corresponding to the candidate interest points from the recommendation degrees of the interest points to be recommended;
acquiring the number of objects to be recommended, and calculating the sum of each candidate recommendation degree corresponding to each candidate interest point to obtain the sum of the candidate recommendation degrees corresponding to each candidate interest point;
performing average calculation based on the sum of the candidate recommendation degrees and the number of the objects to be recommended to obtain the average candidate recommendation degree corresponding to each candidate interest point;
and sorting the average candidate recommendation degrees corresponding to the candidate interest points, and selecting a preset number of candidate interest points as target recommendation interest points according to a sorting result.
9. The method of claim 8, wherein calculating a sum of the candidate recommenders for each candidate interest point to obtain a sum of the candidate recommenders for each candidate interest point comprises:
acquiring recommendation weights corresponding to the objects to be recommended, and performing weighted calculation on the candidate recommendation degrees corresponding to the objects to be recommended according to the recommendation weights to obtain the weighted candidate recommendation degrees;
and calculating the sum of the weighted candidate recommendation degrees of each candidate interest point to obtain the candidate recommendation degree sum of each candidate interest point.
10. The method according to claim 9, wherein the obtaining of the recommendation weight corresponding to each object to be recommended includes:
and obtaining voting information of each object to be recommended on historical recommendation interest points, and calculating recommendation weights corresponding to the objects to be recommended based on the voting information.
11. The method according to claim 8, wherein the step of ranking the average candidate recommendation degrees corresponding to the candidate interest points and selecting a preset number of candidate interest points as target recommendation interest points according to the ranking result comprises:
sorting the average candidate recommendation degrees corresponding to the candidate interest points to obtain a first sorting result;
respectively calculating the distance sum of each candidate interest point to each initial position, and sequencing each candidate interest point according to the distance sum to obtain a second sequencing result;
determining a final recommended point of interest from the candidate points of interest based on the first and second ranking results.
12. The method according to claim 1, wherein the determining a target recommendation interest point based on the interest point attribute information and the recommendation degree of each interest point to be recommended comprises:
determining candidate interest points from the interest points to be recommended based on the interest point attribute information corresponding to the objects to be recommended, and searching candidate recommendation degrees corresponding to the candidate interest points from the recommendation degrees of the interest points to be recommended;
obtaining voting information of each object to be recommended to historical recommendation interest points, and extracting voting characteristics based on the voting information;
calculating the distance sum of each candidate interest point to each starting position, and obtaining distance features based on the distance sum;
acquiring the number of objects to be recommended, and calculating the sum of each candidate recommendation degree corresponding to each candidate interest point to obtain the sum of the candidate recommendation degrees corresponding to each candidate interest point;
performing average calculation based on the sum of the candidate recommendation degrees and the number of the objects to be recommended to obtain an average candidate recommendation degree corresponding to each candidate interest point, and obtaining recommendation degree characteristics based on the average candidate recommendation degree;
inputting the voting features, the distance features and the recommendation degree features into a first interest point recommendation model to obtain target candidate recommendation degrees corresponding to the candidate interest points, and determining model recommendation interest points from the candidate interest points based on the target candidate recommendation degrees, wherein the first interest point recommendation model is obtained by training through training data based on a machine learning algorithm, and the training data comprises a training initial position, training interest point attribute information and training recommendation interest point labels.
13. The method according to any one of claims 8-12, further comprising:
inputting the candidate interest points into a second interest point recommendation model to obtain output target recommendation interest points, wherein the second interest point recommendation model is obtained by training through history candidate interest points and corresponding history recommendation interest points based on a deep neural network algorithm, and the history candidate interest points are obtained by calculating through history initial positions and history interest point attribute information corresponding to history objects to be recommended.
14. A method for showing points of interest, the method comprising:
responding to an interest point recommending event triggered by a target object to be recommended, and displaying an interest point recommending interface, wherein the interest point recommending interface displays an initial position corresponding to each object to be recommended, and each object to be recommended comprises the target object to be recommended;
responding to an interest point attribute information selection event triggered by the interest point recommendation interface, and displaying an interest point attribute information acquisition interface;
responding to an interest point attribute information confirmation event triggered by the interest point attribute information acquisition interface, acquiring target interest point attribute information corresponding to the target object to be recommended, and sending the target interest point attribute information to a server;
displaying target recommendation interest points returned by the server, wherein the target recommendation interest points are candidate interest points determined by the server according to the initial positions corresponding to the objects to be recommended and are determined from the candidate range of the interest points to be recommended based on the attribute information of the interest points corresponding to the objects to be recommended, and is obtained based on the candidate recommendation degrees corresponding to the candidate interest points, which are searched from the target interest point recommendation degree matrix, the target interest point recommendation degree matrix is obtained by performing matrix decomposition prediction by using an original interest point recommendation degree matrix, the original interest point recommendation degree matrix is established by using each interest point to be recommended, the historical interest points and the historical recommendation degree of each object to be recommended corresponding to the historical interest points.
15. An apparatus for point of interest recommendation, the apparatus comprising:
the starting range determining module is used for acquiring starting positions and interest point attribute information corresponding to the objects to be recommended, screening the starting positions to obtain reference starting positions, and determining a starting range based on the reference starting positions;
a candidate range obtaining module, configured to determine a corresponding reference range based on each reference starting position, and obtain a candidate range of interest points to be recommended according to the reference range and the starting range, where the candidate range of interest points to be recommended includes each interest point to be recommended;
an original matrix obtaining module, configured to obtain an original interest point recommendation degree matrix corresponding to each interest point to be recommended, where the original interest point recommendation degree matrix is established using the interest points to be recommended, historical interest points, and historical recommendation degrees of each object to be recommended corresponding to the historical interest points;
the matrix decomposition module is used for carrying out matrix decomposition prediction based on the original interest point recommendation degree matrix to obtain a target interest point recommendation degree matrix, and the target interest point recommendation degree matrix comprises the recommendation degree of each object to be recommended to each interest point to be recommended;
and the interest point recommending module is used for determining a target recommending interest point based on the interest point attribute information corresponding to each object to be recommended and the recommendation degree of each interest point to be recommended and sending the target recommending interest point to the terminal corresponding to each object to be recommended.
16. The apparatus of claim 15, wherein the starting range determining module comprises:
a reference position determining unit, configured to determine a first reference starting position and a second reference starting position based on the position coordinates corresponding to the respective starting positions;
the dividing unit is used for determining a first straight line based on the first reference starting position and the second reference starting position, and dividing each starting position based on the first straight line to obtain a first part starting position and a second part starting position;
the screening unit is used for screening from the initial positions of the first parts to obtain reference initial positions of the first parts and screening from the initial positions of the second parts to obtain reference initial positions of the second parts;
a position obtaining unit, configured to obtain each reference starting position based on the first reference starting position, the second reference starting position, each first partial reference starting position, and each second partial reference starting position, and determine a starting range based on each reference starting position.
17. The apparatus of claim 16, wherein the filtering unit is further configured to calculate a vertical distance between each of the first start positions of the first portion and the first line, and determine a third reference start position from the first start positions of the first portion based on the vertical distance; determining a second straight line based on the third reference starting position and the first reference starting position, dividing the first part starting position based on the second straight line to obtain a third part starting position, determining a third straight line based on the third reference starting position and the second reference starting position, and dividing the first part starting position based on the third straight line to obtain a fourth part starting position; and respectively taking the third part starting position and the fourth part starting position as first part starting positions, respectively taking the second straight line and the third straight line as first straight lines, returning and calculating the vertical distance between each starting position in the first part starting positions and the first straight lines, and executing the step of determining a third reference starting position from the first part starting positions based on the vertical distance until the first part starting positions do not contain the starting positions to obtain each first part reference starting position.
18. The apparatus of claim 15, wherein the candidate range derivation module is further configured to determine each reference range based on each reference starting position and a preset circle radius; and combining the reference ranges and the initial range to obtain the candidate range of the interest point to be recommended.
19. The apparatus according to claim 15, wherein the original matrix obtaining module is further configured to obtain a historical recommendation degree of a historical interest point corresponding to each object to be recommended, obtain a matrix column attribute based on each interest point to be recommended and the historical interest point, and obtain a matrix row attribute based on each object to be recommended; and taking the historical recommendation degree of the historical interest points as a matrix value, and obtaining the original interest point recommendation degree matrix based on the matrix column attribute, the matrix row attribute and the matrix value.
20. The apparatus of claim 15, wherein the matrix factorization module comprises:
the decomposition unit is used for performing matrix decomposition on the original interest point recommendation degree matrix according to a preset decomposition relationship to obtain a first decomposition matrix and a second decomposition matrix, wherein the first decomposition matrix is used for representing the recommendation degree relationship between each object to be recommended and each attribute in the interest point attribute information, and the second decomposition matrix is used for representing the weight relationship between each attribute in the interest point attribute information and the interest point;
the reconstruction unit is used for calculating the product of the first decomposition matrix and the second decomposition matrix to obtain a reconstructed interest point recommendation degree matrix;
the loss calculation unit is used for calculating loss information of the original interest point recommendation degree matrix and the reconstructed interest point recommendation degree matrix and updating the first decomposition matrix and the second decomposition matrix based on the loss information;
the iteration unit is used for returning the step of calculating the product of the first decomposition matrix and the second decomposition matrix to obtain a reconstructed interest point recommendation degree matrix, and performing iteration until the target first decomposition matrix and the target second decomposition matrix are obtained when the loss information meets the preset loss condition;
and the target matrix obtaining unit is used for calculating the product of the target first decomposition matrix and the target second decomposition matrix to obtain the target interest point recommendation degree matrix.
21. The apparatus according to claim 20, wherein the loss calculating unit is further configured to calculate a sum of squared errors of each recommendation in the original point of interest recommendation matrix and a corresponding reconstructed recommendation in the reconstructed point of interest recommendation matrix; calculating first penalty item information corresponding to the first decomposition matrix based on a preset first penalty coefficient, and calculating second penalty item information corresponding to the second decomposition matrix based on a preset second penalty coefficient; and obtaining the loss information based on the error square sum, the first penalty item information and the second penalty item information.
22. The apparatus according to claim 15, wherein the interest point recommending module is further configured to determine candidate interest points from the interest points to be recommended based on the interest point attribute information corresponding to the objects to be recommended, and search for candidate recommendation degrees corresponding to the candidate interest points from the recommendation degrees of the interest points to be recommended; acquiring the number of objects to be recommended, and calculating the sum of each candidate recommendation degree corresponding to each candidate interest point to obtain the sum of the candidate recommendation degrees corresponding to each candidate interest point; performing average calculation based on the sum of the candidate recommendation degrees and the number of the objects to be recommended to obtain the average candidate recommendation degree corresponding to each candidate interest point; and sorting the average candidate recommendation degrees corresponding to the candidate interest points, and selecting a preset number of candidate interest points as target recommendation interest points according to a sorting result.
23. The apparatus according to claim 22, wherein the interest point recommending module is further configured to obtain a recommendation weight corresponding to each object to be recommended, and perform weighted calculation on the candidate recommendation degrees corresponding to each object to be recommended according to the recommendation weight to obtain each weighted candidate recommendation degree; and calculating the sum of the weighted candidate recommendation degrees of each candidate interest point to obtain the candidate recommendation degree sum of each candidate interest point.
24. The apparatus of claim 23, wherein the interest point recommending module is further configured to obtain voting information of each object to be recommended for a historical recommendation interest point, and calculate a recommendation weight corresponding to each object to be recommended based on the voting information.
25. The apparatus according to claim 22, wherein the interest point recommending module is further configured to rank the average candidate recommendation degrees corresponding to the candidate interest points to obtain a first ranking result; respectively calculating the distance sum of each candidate interest point to each initial position, and sequencing each candidate interest point according to the distance sum to obtain a second sequencing result; determining a final recommended point of interest from the candidate points of interest based on the first and second ranking results.
26. The apparatus according to claim 15, wherein the interest point recommending module is further configured to determine candidate interest points from the interest points to be recommended based on the interest point attribute information corresponding to the objects to be recommended, and search for candidate recommendation degrees corresponding to the candidate interest points from the recommendation degrees of the interest points to be recommended; obtaining voting information of each object to be recommended to historical recommendation interest points, and extracting voting characteristics based on the voting information; calculating the distance sum of each candidate interest point to each starting position, and obtaining distance features based on the distance sum; acquiring the number of objects to be recommended, and calculating the sum of each candidate recommendation degree corresponding to each candidate interest point to obtain the sum of the candidate recommendation degrees corresponding to each candidate interest point; performing average calculation based on the sum of the candidate recommendation degrees and the number of the objects to be recommended to obtain an average candidate recommendation degree corresponding to each candidate interest point, and obtaining recommendation degree characteristics based on the average candidate recommendation degree; inputting the voting features, the distance features and the recommendation degree features into a first interest point recommendation model to obtain target candidate recommendation degrees corresponding to the candidate interest points, and determining model recommendation interest points from the candidate interest points based on the target candidate recommendation degrees, wherein the first interest point recommendation model is obtained by training through training data based on a machine learning algorithm, and the training data comprises a training initial position, training interest point attribute information and training recommendation interest point labels.
27. The apparatus of any one of claims 22-26, further comprising:
and the model recommendation module is used for inputting the candidate interest points into a second interest point recommendation model to obtain output target recommendation interest points, the second interest point recommendation model is obtained by training through historical candidate interest points and corresponding historical recommendation interest points based on a deep neural network algorithm, and the historical candidate interest points are obtained by calculating through historical initial positions and historical interest point attribute information corresponding to the historical objects to be recommended.
28. A point of interest presentation apparatus, the apparatus comprising:
the recommendation interface display module is used for responding to an interest point recommendation event triggered by a target object to be recommended and displaying an interest point recommendation interface, wherein the interest point recommendation interface displays an initial position corresponding to each object to be recommended, and each object to be recommended comprises the target object to be recommended;
the information acquisition interface display module is used for responding to an interest point attribute information selection event triggered by the interest point recommendation interface and displaying an interest point attribute information acquisition interface;
the information sending module is used for responding to an interest point attribute information confirmation event triggered by the interest point attribute information acquisition interface, acquiring target interest point attribute information corresponding to the target object to be recommended and sending the target interest point attribute information to a server;
an interest point display module, configured to display a target recommended interest point returned by the server, where the target recommended interest point is a candidate range of interest points to be recommended that is determined by the server according to an initial position corresponding to each object to be recommended, and each candidate interest point is determined from the candidate range of interest points to be recommended based on interest point attribute information corresponding to each object to be recommended, and is obtained based on the candidate recommendation degrees corresponding to the candidate interest points, which are searched from the target interest point recommendation degree matrix, the target interest point recommendation degree matrix is obtained by performing matrix decomposition prediction by using an original interest point recommendation degree matrix, the original interest point recommendation degree matrix is established by using each interest point to be recommended, the historical interest points and the historical recommendation degree of each object to be recommended corresponding to the historical interest points.
29. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the method of any of claims 1 to 14.
30. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 14.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110054033.4A CN112711713B (en) | 2021-01-15 | 2021-01-15 | Interest point recommendation and display method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110054033.4A CN112711713B (en) | 2021-01-15 | 2021-01-15 | Interest point recommendation and display method and device, computer equipment and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112711713A CN112711713A (en) | 2021-04-27 |
CN112711713B true CN112711713B (en) | 2022-04-26 |
Family
ID=75549094
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110054033.4A Active CN112711713B (en) | 2021-01-15 | 2021-01-15 | Interest point recommendation and display method and device, computer equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112711713B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113254811B (en) * | 2021-06-23 | 2021-09-24 | 佛山市墨纳森智能科技有限公司 | Interest place recommendation method, system and storage medium based on big data |
CN114372213B (en) * | 2022-03-22 | 2022-06-10 | 腾讯科技(深圳)有限公司 | Information processing method, apparatus, device, storage medium, and program product |
CN117216397B (en) * | 2023-09-22 | 2024-03-12 | 哈尔滨工业大学 | POI recommendation method based on regional influence depth matrix decomposition |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112214677A (en) * | 2020-10-19 | 2021-01-12 | 腾讯科技(深圳)有限公司 | Interest point recommendation method and device, electronic equipment and storage medium |
Family Cites Families (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8145661B1 (en) * | 2008-03-31 | 2012-03-27 | Unites Services Automobile Association (USAA) | Systems and methods for a home area recommender |
CN103023977B (en) * | 2012-11-19 | 2015-07-01 | 华南理工大学 | Recommendation system and method based on positional information |
CN104112213B (en) * | 2013-04-19 | 2018-09-04 | 腾讯科技(深圳)有限公司 | The method and device of recommendation information |
GB201506356D0 (en) * | 2015-04-15 | 2015-05-27 | Tomtom Int Bv | Methods of obtaining point of interest data |
CN106844376B (en) * | 2015-12-03 | 2020-03-03 | 高德软件有限公司 | Method and device for recommending interest points |
CN106776993B (en) * | 2016-12-06 | 2020-07-24 | 苏州大学 | Path recommendation method and system based on time sequence constraint activity intention |
CN111831899B (en) * | 2019-09-24 | 2023-11-24 | 北京嘀嘀无限科技发展有限公司 | Navigation interest point recommendation method, device, server and readable storage medium |
CN111680228B (en) * | 2020-06-11 | 2022-03-18 | 浙江工商大学 | Interest point recommendation method based on geographic position fusion and category popularity |
-
2021
- 2021-01-15 CN CN202110054033.4A patent/CN112711713B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112214677A (en) * | 2020-10-19 | 2021-01-12 | 腾讯科技(深圳)有限公司 | Interest point recommendation method and device, electronic equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN112711713A (en) | 2021-04-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11238473B2 (en) | Inferring consumer affinities based on shopping behaviors with unsupervised machine learning models | |
CN112711713B (en) | Interest point recommendation and display method and device, computer equipment and storage medium | |
US11270324B2 (en) | Location-based analytic platform and methods | |
CN111008332B (en) | Content item recommendation method, device, server and storage medium | |
US10198635B2 (en) | Systems and methods for associating an image with a business venue by using visually-relevant and business-aware semantics | |
US10176244B2 (en) | Text characterization of trajectories | |
CN107424043A (en) | A kind of Products Show method and device, electronic equipment | |
CN110503506B (en) | Item recommendation method, device and medium based on grading data | |
CN108897750B (en) | Personalized place recommendation method and device integrating multiple contextual information | |
US20120296776A1 (en) | Adaptive interactive search | |
CN112131261B (en) | Community query method and device based on community network and computer equipment | |
CN113627995B (en) | Commodity recommendation list updating method and device, equipment, medium and product thereof | |
CN110750697B (en) | Merchant classification method, device, equipment and storage medium | |
JP2018128942A (en) | Analyzing apparatus, analyzing method, and program | |
CN110598084A (en) | Object sorting method, commodity sorting device and electronic equipment | |
CN115131088A (en) | Intelligent computer search function for locating items of interest near a user | |
Lin et al. | Extracting urban landmarks from geographical datasets using a random forests classifier | |
CN111563207B (en) | Search result sorting method and device, storage medium and computer equipment | |
CN117709968A (en) | Personnel allocation method, apparatus, computer device and storage medium | |
CN116485503A (en) | Commodity combination recommendation method, device, equipment and medium thereof | |
CN115758271A (en) | Data processing method, data processing device, computer equipment and storage medium | |
CN113704617A (en) | Article recommendation method, system, electronic device and storage medium | |
CN111291829A (en) | Automatic determination method and system for selected pictures | |
CN112882621B (en) | Module display method, module display device, computer equipment and storage medium | |
Wang et al. | A feature extraction and similarity metric-learning framework for urban model retrieval |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 40042952 Country of ref document: HK |
|
GR01 | Patent grant | ||
GR01 | Patent grant |