CN110929162A

CN110929162A - Recommendation method and device based on interest points, computer equipment and storage medium

Info

Publication number: CN110929162A
Application number: CN201911227981.2A
Authority: CN
Inventors: 邓颖; 张金超; 牛成
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2019-12-04
Filing date: 2019-12-04
Publication date: 2020-03-27
Anticipated expiration: 2039-12-04
Also published as: CN110929162B

Abstract

The invention discloses a recommendation method, a recommendation device, computer equipment and a storage medium based on interest points, which can determine geographical interest points of historical sign-in of a user based on historical sign-in data and generate a geographical interest point access sequence based on access relations among the geographical interest points; the method comprises the steps of learning an access rule of the geographic interest points in a geographic interest point access sequence based on the obtained description vector of the geographic interest points, obtaining access relation vectors corresponding to the geographic interest points, and generating relevant recommendation information of the geographic interest points based on the access relation vectors.

Description

Recommendation method and device based on interest points, computer equipment and storage medium

Technical Field

The invention relates to the technical field of internet, in particular to a recommendation method and device based on points of interest, computer equipment and a storage medium.

Background

At present, with the rapid development and popularization of mobile internet technology, more and more users are used to determine various purchasing behaviors through information provided by the internet, and a location-based internet business model is also developed.

A Point of Interest (POI) is a term in a geographic information system, generally referring to all geographic objects that can be abstracted as points. At present, the association between geographic interest points can be researched based on historical check-in data of users, but in the related art, generally, the historical check-in sequence of each user is taken as a research object, and the check-in data of the users is sparse, so that the association between a plurality of geographic interest points is difficult to establish, and the geographic interest points are not beneficial to the recommendation service about the position.

Disclosure of Invention

The embodiment of the invention provides a recommendation method and device based on interest points, computer equipment and a storage medium, which are beneficial to deeply mining the association relation between geographic interest points and improving the recommendation accuracy based on the interest points.

The embodiment of the invention provides a recommendation method based on interest points, which comprises the following steps:

acquiring historical check-in data of a user, and determining geographic interest points of the historical check-in of the user from the historical check-in data;

acquiring access relation information among the geographic interest points based on the historical check-in data;

generating a geographical interest point access sequence based on the access relationship information among the geographical interest points, wherein adjacent geographical interest points in the geographical interest point access sequence are accessed by at least one user successively;

obtaining a description vector of the geographic interest point;

learning an access rule of the geographic interest points in the geographic interest point access sequence based on the description vector of the geographic interest points to obtain access relation vectors corresponding to the geographic interest points, wherein the access relation vectors are used for representing the association of the corresponding geographic interest points and other geographic interest points on user access behaviors;

and generating relevant recommendation information of the geographic interest points based on the access relation vectors of the geographic interest points.

Optionally, the obtaining the description vector of the geographic interest point includes:

acquiring user check-in texts of the geographic interest points from the historical check-in data;

semantic analysis is carried out on the user check-in text of the geographic interest points to obtain text description vectors used for describing the semantics of the user check-in text, wherein the text description vectors are the same in length;

and fusing all the text description vectors of the same geographic interest point to obtain a fused vector, and taking the fused vector as the description vector of the corresponding geographic interest point.

Optionally, the fusing all text description vectors of the same geographic interest point to obtain a fused vector, and taking the fused backward vector as a description vector of the corresponding geographic interest point includes:

and averaging all the text description vectors of the same geographic interest point to obtain an average vector as the description vector of the corresponding geographic interest point.

Optionally, learning an access rule of the geographic interest points in the geographic interest point access sequence based on the description vector of the geographic interest points to obtain an access relationship vector corresponding to the geographic interest points, including:

learning the access rule of the geographic interest points in the geographic interest point access sequence based on the description vector of the geographic interest points through an access rule analysis model to obtain access relation vectors corresponding to the geographic interest points;

before the learning of the access rule of the geographic interest points in the geographic interest point access sequence based on the description vector of the geographic interest points to obtain the access relationship vector corresponding to the geographic interest points, the method further comprises the following steps:

acquiring unique hot codes of the geographic interest points, and arranging the unique hot codes of the geographic interest points according to the arrangement sequence of the geographic interest points in the geographic interest point access sequence to obtain unique hot code sequences corresponding to the geographic interest points;

and training the access law analysis model through the unique hot code sequence so that the access law analysis model learns the access law of the geographic interest points in the access sequence of the geographic interest points.

Optionally, the obtaining access relationship information between the geographic points of interest based on the historical check-in data includes:

determining a user access direction between every two geographic interest points and an access preference degree corresponding to the user access direction based on the historical check-in data;

drawing a directed weighted graph of the geographic interest points based on the user access directions among the geographic interest points and the access preference degrees corresponding to the user access directions, wherein nodes in the directed weighted graph represent the geographic interest points, directed edges in the directed weighted graph represent the user access directions between the two geographic interest points, and the weights of the directed edges are obtained based on the access preference degrees corresponding to the user access directions;

the generating of the access sequence of the geographic interest points based on the access relation information among the geographic interest points comprises:

and generating a geographic interest point access sequence with a preset length based on the nodes in the directed weighted graph and the weights of the directed edges.

Optionally, the generating a geographic interest point access sequence with a preset length based on the weights of the nodes and the directed edges in the directed weighted graph includes:

selecting a part of nodes from the directed weighted graph as starting nodes of a geographic interest point access sequence;

based on the weight of each directed edge in the directed weighted graph, starting wandering from each initial node in the directed weighted graph, and generating a geographic interest point access sequence based on nodes passing through a wandering path, wherein the length of the geographic interest point access sequence is a preset length, and adjacent nodes of the same node are different in the wandering path.

Optionally, the user preference degree includes a user access number;

the method for drawing the directional weighted graph of the geographic interest points based on the user access direction among the geographic interest points and the access preference degree corresponding to the user access direction comprises the following steps:

generating nodes of a directed weighted graph based on the geographic interest points, and generating directed edges among the nodes based on the user access directions among the geographic interest points;

calculating the user access times and the total user access times corresponding to a first directed edge pointing to other geographic interest points from the same geographic interest point in the directed weighted graph;

and calculating the ratio of the user access times of the connected first directed edges to the total corresponding user access times of the same geographic interest point, and taking the ratio as the weight of the corresponding first directed edge.

Optionally, the learning the access rule of the geographic interest point in the geographic interest point access sequence based on the description vector of the geographic interest point to obtain an access relationship vector corresponding to the geographic interest point, further includes:

and fusing the description vector corresponding to the geographic interest point and the access relation vector to obtain a comprehensive description vector of the geographic interest point.

The embodiment also provides a recommendation device based on the point of interest, which includes:

the check-in data acquisition unit is used for acquiring historical check-in data of the user and determining the geographic interest points of the historical check-in of the user from the historical check-in data;

the access relation obtaining unit is used for obtaining access relation information among the geographic interest points based on the historical sign-in data;

the sequence generating unit is used for generating a geographical interest point access sequence based on the access relation information among the geographical interest points, wherein the adjacent geographical interest points in the geographical interest point access sequence are accessed by at least one user successively;

the description vector acquisition unit is used for acquiring the description vector of the geographic interest point;

the access relation vector acquisition unit is used for learning the access rules of the geographic interest points in the geographic interest point access sequence based on the description vectors of the geographic interest points to obtain access relation vectors corresponding to the geographic interest points, wherein the access relation vectors are used for representing the association of the corresponding geographic interest points and other geographic interest points on user access behaviors;

and the recommending unit is used for generating the relevant recommending information of the geographic interest points based on the access relation vectors of the geographic interest points.

Optionally, the description vector obtaining unit includes:

the text acquisition subunit is used for acquiring the user check-in text of the geographic interest point from the historical check-in data;

the semantic analysis subunit is used for performing semantic analysis on the user check-in text of the geographic interest point to obtain text description vectors for describing the semantics of the user check-in text, wherein the text description vectors are the same in length;

and the fusion subunit is used for fusing all the text description vectors of the same geographic interest point to obtain a fused vector, and taking the fused backward quantity as the description vector of the corresponding geographic interest point.

Optionally, the merging subunit is configured to average all text description vectors of the same geographic interest point, and obtain an average vector as a description vector of the corresponding geographic interest point.

Optionally, the access relationship vector obtaining unit is configured to learn, through the access rule analysis model and based on the description vector of the geographic interest point, an access rule of the geographic interest point in the geographic interest point access sequence to obtain an access relationship vector corresponding to the geographic interest point;

the apparatus of this embodiment further comprises: the first model training unit is used for acquiring the unique hot codes of the geographic interest points before the access relation vector acquisition unit learns the access rules of the geographic interest points in the access sequence of the geographic interest points through the access rule analysis model and based on the description vectors of the geographic interest points to obtain the access relation vectors corresponding to the geographic interest points, and arranging the unique hot codes of the geographic interest points according to the arrangement sequence of the geographic interest points in the access sequence of the geographic interest points to obtain the unique hot code sequences corresponding to the geographic interest points; and training the access law analysis model through the unique hot code sequence so that the access law analysis model learns the access law of the geographic interest points in the access sequence of the geographic interest points.

Optionally, the access relationship obtaining unit includes:

the access information determining subunit is used for determining a user access direction between every two geographic interest points and an access preference degree corresponding to the user access direction based on the historical check-in data;

the directed weighted graph generating subunit is used for drawing a directed weighted graph of the geographic interest points based on the user access direction between the geographic interest points and the access preference degree corresponding to the user access direction, wherein nodes in the directed weighted graph represent the geographic interest points, directed edges in the directed weighted graph represent the user access direction between the two geographic interest points, and the weight of the directed edges is obtained based on the access preference degree corresponding to the user access direction;

correspondingly, the sequence generating unit is configured to generate a geographic interest point access sequence with a preset length based on the nodes in the directed weighted graph and the weights of the directed edges.

Optionally, the sequence generating unit includes:

the selecting subunit is used for selecting a part of nodes from the directed weighted graph as initial nodes of the geographic interest point access sequence;

and the generating subunit is configured to, based on the weight of each directed edge in the directed weighted graph, start walking from each start node in the directed weighted graph, and generate a geographic interest point access sequence based on nodes that pass through a walking path, where a length of the geographic interest point access sequence is a preset length, and in the walking path, adjacent nodes of the same node are different.

Optionally, the user preference degree includes a user access number;

a generating subunit for:

Optionally, the apparatus of this embodiment further includes: and the fusion unit is used for learning the access rule of the geographic interest points in the geographic interest point access sequence through the access rule analysis model based on the description vectors of the geographic interest points in the access relation vector acquisition unit to obtain the access relation vectors corresponding to the geographic interest points, and then fusing the description vectors corresponding to the geographic interest points with the access relation vectors to obtain the comprehensive description vectors of the geographic interest points.

The present embodiment also provides a storage medium having a computer program stored thereon, wherein the computer program, when being executed by a processor, implements the steps of the point of interest based recommendation method as described above.

The present embodiment also provides a computer device, including a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the point of interest-based recommendation method as described above.

The embodiment discloses a recommendation method and device based on interest points, computer equipment and a storage medium, which can acquire historical sign-in data of a user and determine geographical interest points of the historical sign-in of the user from the historical sign-in data; acquiring access relation information between geographic interest points based on historical sign-in data; generating a geographical interest point access sequence based on access relation information among the geographical interest points, wherein adjacent geographical interest points in the geographical interest point access sequence are accessed by at least one user successively; obtaining description vectors of geographic interest points; learning the access rules of the geographic interest points in the geographic interest point access sequence based on the description vectors of the geographic interest points to obtain access relationship vectors corresponding to the geographic interest points, wherein the access relationship vectors are used for representing the association of the corresponding geographic interest points and other geographic interest points on user access behaviors, and generating relevant recommendation information of the geographic interest points based on the access relationship vectors of the geographic interest points. And determining appropriate recommendation information, thereby improving the accuracy and reliability of recommendation services based on the association.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

FIG. 1a is a scene diagram of a point of interest-based recommendation method according to an embodiment of the present invention;

FIG. 1b is a flowchart of a point of interest-based recommendation method according to an embodiment of the present invention;

FIG. 2a is a schematic diagram of a directed weighted graph in accordance with an embodiment of the present invention;

FIG. 2b is a schematic diagram of the training of the dialog model and the skip-gram model in an embodiment of the present invention;

FIG. 3 is a schematic structural diagram of a point of interest-based recommendation apparatus according to an embodiment of the present invention;

FIG. 4 is a schematic structural diagram of a computer device provided by an embodiment of the present invention;

fig. 5 is an alternative structure diagram of the distributed system 100 applied to the blockchain system according to the embodiment of the present invention;

fig. 6 is an alternative schematic diagram of a block structure according to an embodiment of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

The embodiment of the invention provides a recommendation method and device based on interest points, computer equipment and a storage medium. Specifically, the embodiment of the present invention provides a point of interest-based recommendation method applicable to a computer device, where the computer device may be a terminal, the terminal may be a mobile phone, a tablet computer, a notebook computer, or another device, or may be a server, and the server may be a single server or a server cluster including multiple servers.

For example, the point of interest-based recommendation apparatus may be integrated in a terminal, or may be integrated in a server.

The embodiment of the invention takes computer equipment as a server as an example to introduce a point-of-interest-based recommendation method.

Referring to fig. 1a, a point of interest-based recommendation system provided by an embodiment of the present invention includes a terminal 10, a server 20, and the like; the terminal 10 and the server 20 are connected through a network, such as a wired or wireless network, and the like, wherein the point-of-interest based recommendation device is integrated in the server.

The server 20 may be configured to obtain historical check-in data of the user, and determine geographic interest points of the historical check-in of the user from the historical check-in data; acquiring access relation information among the geographic interest points based on the historical check-in data; generating a geographical interest point access sequence based on the access relationship information among the geographical interest points, wherein adjacent geographical interest points in the geographical interest point access sequence are accessed by at least one user successively; obtaining a description vector of the geographic interest point; learning the access rule of the geographic interest points in the geographic interest point access sequence based on the description vector of the geographic interest points to obtain access relation vectors corresponding to the geographic interest points, and generating relevant recommendation information of the geographic interest points based on the access relation vectors, wherein the access relation vectors are used for representing the relevance of the corresponding geographic interest points and other geographic interest points on user access behaviors.

The visiting relation vector of the geographic interest point may specifically include information of geographic interest points visited by the user before and/or after the geographic interest point, and user visiting probabilities of the geographic interest points.

Specifically, the server 20 is further configured to, when receiving a recommendation request for a target geographic interest point, obtain an access relationship vector corresponding to the target geographic interest point; and determining recommended geographic interest points meeting the recommendation purposes of the recommendation requests based on the access relation vectors, and generating recommendation results responding to the recommendation requests based on the recommended geographic interest points.

The recommendation request may be sent by the user through the terminal.

Optionally, the terminal 10 may be configured to determine a target geographic interest point where the terminal is currently located when detecting a location-based service triggered by a user, generate a recommendation request, and send the recommendation request to the server, where the recommendation request carries information of the geographic interest point where the terminal is currently located and a recommendation purpose of the recommendation request.

The server 20 may be configured to receive the recommendation request, obtain information of the target geographic interest point therein, obtain an access relationship vector of the target geographic interest point based on the information, determine a recommendation result fed back to the terminal based on the access relationship vector and a recommendation purpose of the recommendation request, and send the recommendation result to the terminal.

The recommendation request may be for requesting the server 20 to recommend the geographic interest point, the server 20 may obtain information of the geographic interest point in the recommendation request, obtain an access relationship vector of the geographic interest point, determine, based on the access relationship vector, the geographic interest point with the highest user access probability after the geographic interest point is determined, use the determined geographic interest point as the geographic interest point recommended to the user, and send the information of the determined geographic interest point to the terminal.

The recommendation request may be for requesting the server 20 to recommend the path, and the server 20 may obtain information of the geographic interest point in the recommendation request, obtain an access relationship vector of the geographic interest point, and determine several subsequent geographic interest points based on the access relationship vector

And the user accesses the sequence and the geographical interest point with the highest user access probability, the determined geographical interest point is used as the geographical interest point recommended to the user, and the information of the determined geographical interest point is sent to the terminal.

The following are detailed below. It should be noted that the following description of the embodiments is not intended to limit the preferred order of the embodiments.

The embodiment of the invention will be described from the perspective of a point-of-interest-based recommendation device, which may be specifically integrated in a terminal or a server.

In the recommendation method based on the point of interest provided in the embodiment of the present invention, the method may be executed by a processor of a server, as shown in fig. 1b, a specific process of the recommendation method based on the point of interest may be as follows:

101. acquiring historical sign-in data of a user, and determining geographic interest points of the historical sign-in of the user from the historical sign-in data;

in this embodiment, the area for acquiring the historical check-in data and the time to which the historical check-in data belongs are not displayed, and may be set according to actual needs.

Optionally, the step of "acquiring historical check-in data of the user" may include: and acquiring historical sign-in data of users in a preset area. The size and the position of the region can be set according to actual needs, for example, the region is divided by a geographic administrative unit to acquire historical sign-in data of users in each geographic administrative unit, or the region is divided by longitude and latitude to acquire historical sign-in data of users in a preset longitude and latitude range.

Optionally, in this embodiment, for accuracy of location services such as location recommendation, route planning, and the like, historical sign-in data of the user may be acquired once per day at time intervals of day, and subsequent steps are performed to update the comprehensive description vector of the geographic point of interest.

Optionally, the step of "acquiring historical check-in data of the user" may include: and when the preset updating time point is reached, acquiring historical sign-in data of the user. Wherein the preset update time point may be set to an arbitrary time point of each day, etc., or the preset update time point may be set to a time point of monday, tuesday, or morning of wednesday, etc.

In this embodiment, the historical check-in data of the user may be check-in data within a preset time period before the obtaining time, where the preset time period may be set as required, for example, set to 7 days, 8 days, and the like, and this embodiment does not limit this.

The geographic point of interest POI in this embodiment may be any geographic entity, especially some geographic entities closely related to life of people, such as schools, banks, restaurants, gas stations, hospitals, supermarkets, scenic spots, shopping malls, and shopping malls. It may contain four basic attributes of name, address, coordinates, and category, and may also contain other attributes such as social functionality. The geographic interest points are important information in the map and have important significance in the aspects of navigation, recommendation and the like. The main purpose of the geographic 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 enhanced to a great extent, and the accuracy and the speed of geographic positioning are improved.

In this embodiment, the historical check-in data of the user includes check-in text of the geographic point of interest, and the check-in text includes types, features, and user experience information of the geographic point of interest, where the types of the geographic point of interest include, for example, school, bank, restaurant, gas station, hospital, supermarket, scenic spot, mall, and the like, and the features of the geographic point of interest, such as features for mall, may include: the size of the mall, the amount of people, the type of the transaction item, and the like, and the user experience information of the geographic interest point may include user comment information of the user on the geographic interest point, and the like. The user comment information can comprise comment texts input by the user, and can also comprise comment scores input by the user in a comment page of the geographic interest points, and the like.

It can be understood that, in the present embodiment, there may be multiple check-ins for multiple users in one geographic point of interest, so that one geographic point of interest may have multiple check-in texts.

102. Acquiring access relation information between geographic interest points based on historical sign-in data;

in this embodiment, the access relationship information between the geographic interest points may be specifically understood as access relationship information between every two geographic interest points in all the geographic interest points, for example, the access relationship information includes: a user visit direction between any two geographic points of interest and a degree of visit preference in the user visit direction.

The user access direction can be understood as the user access sequence between two geographic interest points. The access preference degree in the user access direction may be determined by the access data in the user access direction, for example, the user access preference degree in the user access direction may be determined based on the number of accesses in each user access direction. Alternatively, the degree of user access preference in the user access direction may be determined based on the frequency of user access in the user access direction over a period of time.

For the scheme of determining the user access preference degree in the user access direction based on the access times in the respective user access directions, it can be understood that the greater the access times, the higher the user access preference degree in the user access direction.

In this embodiment, the obtained access relationship information may be obtained by analyzing two geographical interest points as a basic unit, the influence of the user dimension on the access relationship between the geographical interest points is weakened through the step 102, the access relationship between the user and the geographical interest points is no longer distinguished from the user dimension, which is equivalent to splitting the sign-in procedure of each user for the geographical interest points into associations between a group of two geographical interest points, and the access relationship information is obtained by analyzing the access relationship between any two geographical interest points of all the users.

Optionally, the step of obtaining access relationship information between geographic points of interest based on historical check-in data may include:

combining every two geographic interest points determined based on historical check-in data to determine all geographic interest point pairs obtained through combination;

and determining access relation information between each geographical interest point pair based on the historical check-in data.

The step of determining access relationship information between each geographical interest point pair based on historical check-in data may include: based on historical check-in data, determining a user access direction corresponding to each geographical interest point and an access preference degree of the user in the user access direction; and taking the user access direction and the corresponding access preference degree as access relation information among the geographic interest points.

For example, taking geographic interest points a and B as an example, assuming that only from historical check-in data of the user 1, the user 1 is found to have the behavior of "check-in first at the geographic interest point a and then check-in again at the geographic interest point B" for 1 time, then a user access direction pointing from a to B exists between the geographic interest points a and B, and the number of accesses in the user access direction is 1, assuming that only from the historical check-in data of user 2, the user 2 is found to have the behavior of checking in at the geographic point of interest B first and then checking in at the geographic point of interest a again 2 times, then there is a user access direction from B to a between the geographic points of interest a and B, and the number of accesses in the user access direction is 2, it can be understood that the user access direction from B to a is greater than the number of accesses in the user access direction from a to B, so the user access direction from B to a has a higher degree of access preference.

In this embodiment, the access relationship information between the geographical interest point pairs may be stored in a table form, for example, after the access relationship information between the geographical interest points is obtained, the access relationship information of all the geographical interest pairs is written into a preset table, and the preset table is stored into the corresponding block chain.

In another embodiment, the visiting relationship information between the geographic interest points can be more intuitively represented in a directional weighted graph mode.

Optionally, the access relationship information between the geographic interest points includes a user access direction between the geographic interest points and an access preference degree corresponding to the user access direction, and the step "obtaining the access relationship information between the geographic interest points based on the historical check-in data" may include:

determining a user access direction between every two geographic interest points and an access preference degree corresponding to the user access direction based on historical check-in data;

and drawing a directed weighted graph of the geographic interest points based on the user access directions among the geographic interest points and the access preference degrees corresponding to the user access directions, wherein nodes in the directed weighted graph represent the geographic interest points, directed edges in the directed weighted graph represent the user access directions between the two geographic interest points, and the weights of the directed edges are obtained based on the access preference degrees corresponding to the user access directions.

In practice, the sequence of the user's visits between two POIs often reflects the influence of factors such as POI popularity, POI geographical location distribution, and traffic convenience, and in order to consider the visit sequence, this embodiment uses each POI as a node in a directed weighted graph, and represents the visit record (i.e., the user visit direction) from one POI to the next POI with a directed edge. In the historical check-in data, the check-in sequences of all users are divided into two adjacent POI- > POI pairs, directional connection between POI nodes is constructed, the weight of each directional edge is obtained based on the access preference degree corresponding to each directional edge, and finally a directional weighted graph is obtained.

In this embodiment, the access preference degree may be calculated based on the number of accesses of the user in the user access direction, or may be directly represented by the number of accesses in the user access direction.

Optionally, the user preference degree includes the number of user accesses; the step of drawing a directed weighted graph of the geographical interest points based on the user access directions among the geographical interest points and the access preference degrees corresponding to the user access directions may include:

generating nodes of the directed weighted graph based on the geographic interest points, and generating directed edges between the nodes of the directed weighted graph based on the user access directions between the geographic interest points;

For example, referring to the directed weighted graph shown in FIG. 2a, there is a node y pointing to all nodes { x }₁,x₂,…,x_nThe nodes pointed to it include z1, etc. Where the directed edges pointing from y to all x may be derived from historical check-in data of different users.

Referring to FIG. 2a, the first directional edge of a geographic point of interest y is pointing from the geographic point of interest y to a point of interest { x₁,x₂,…,x_nN directed edges (the n directed edges are not completely shown in fig. 2 a), and a weight is correspondingly set on each first directed edge, where the weight may reflect the access preference degree of the user on the directed edge.

Suppose geographic point of interest y points to geographic point of interest { x₁,x₂,…,x_nThe times (i.e. the user access times corresponding to each first directed edge) are sequentially { c }₁,c₂,…,c_nPointing the geographic interest point y to the geographic interest point { x }₁,x₂,…,x_nThe weight of the first directed edge of the lattice is in turn

For each node in the directed weighted graph, the weight of each first directed edge of the connection of that node can be determined in the manner described above, e.g., for z in FIG. 2a₁The node can calculate the node z in the directed weighted graph by the scheme₁To other nodes, e.g. node z₁The weight of the first directed edge pointing to node y.

103. And generating a geographic interest point access sequence based on the access relation information among the geographic interest points, wherein the adjacent geographic interest points in the geographic interest point access sequence are accessed by at least one user successively.

Optionally, the length of the geographical interest point visit sequence in this embodiment may be the same, that is, the total number of geographical interest points in each geographical interest point visit sequence is the same.

In this embodiment, the geographic interest point visiting sequence may be determined based on a user visiting direction between geographic interest points and a visiting preference degree of the user in the user visiting direction. And each geographic interest point visit sequence has a visit relationship between adjacent geographic interest points, namely the adjacent geographic interest points in the sequence are visited by at least one user successively.

For example, the sequence of geographic interest Q is Q1, Q2, Q3, Q4,. cndot.. Qm. At least one user checks in at geographic point of interest Q1 first, and then checks in at geographic point of interest Q2. In this embodiment, in the same geographic interest point sequence, the visiting users may be different for different groups of adjacent geographic interest points, for example, the geographic interest points Q1 and Q2 are visited by the user C sequentially, and the geographic interest points Q2 and Q3 are visited by the users D and E sequentially.

Optionally, the step "generating a geographic interest point visit sequence based on the visit relationship information between geographic interest points" may include:

and generating a geographic interest point visit sequence based on the visit direction of the users among the geographic interest points and the visit preference degree in each visit direction of the users.

In this embodiment, the number of the generated geographic interest point access sequences may be preset, and this embodiment does not limit this.

In one embodiment, the geographic points of interest determined based on historical check-in data occur at least once all in the sequence of geographic point of interest visits.

Further, the geographic point of interest determined based on the historical check-in data appears at least once as a non-starting point and an ending point in the sequence of geographic point of interest visits.

Optionally, the step "generating a geographic interest point visit sequence based on the user visit directions between the geographic interest points and the visit preference degrees in the respective user visit directions" may include:

selecting a part from the geographic interest points determined based on the historical check-in data as a starting node of the geographic interest point access sequence;

for each geographic interest point serving as a starting node, selecting one of other geographic interest points as a next geographic interest point in a geographic interest point access sequence based on a first user access direction and an access preference degree in the first user access direction of the geographic interest point and the other geographic interest points, wherein the first access direction is an access direction in which the geographic interest point of the starting node points to the other geographic interest points;

and taking the selected geographic interest points as starting nodes, and returning to execute the step of selecting one of the other geographic interest points as a next geographic interest point in the geographic interest point visit sequence for each geographic interest point serving as the starting node based on the first user visit direction and the visit preference degree in the first user visit direction of the geographic interest point and the other geographic interest points until the length of the geographic interest point visit sequence is the preset length.

For each geographic interest point serving as a starting node, the other geographic interest points refer to geographic interest points other than the geographic interest point of the starting node.

In this embodiment, for the access relationship information between the geographic interest points, through a scheme embodied by a directed weighted graph similar to that shown in fig. 2a, the step "generating the access sequence of the geographic interest points based on the access relationship information between the geographic interest points" may include:

Optionally, the step "generating a geographic interest point access sequence with a preset length based on the weights of the nodes and the directed edges in the directed weighted graph" may include:

selecting a part of nodes from the directed weighted graph as initial nodes of a geographic interest point access sequence;

In this embodiment, the starting node of the geographic interest point access sequence may be selected from all geographic interest points in a sampling manner, and the walking manner in this embodiment may be random walking.

For example, in this embodiment, a batch of nodes may be uniformly sampled among all the geographic interest points as the starting node of the geographic interest point visit sequence. Then, based on the directed weighted graph, a fixed length, such as L, of the geographical interest point visit sequence is generated by a weighted walking mode from the starting node to the node. In this embodiment, it is considered that if the geographic interest points in the geographic interest point access sequence form a small closed loop, the data validity of the sequence may be reduced, so that the previous node will not be used as a next sampling target in the walking process on the directed weighted graph. I.e., the walk path, the neighbors of each node are different.

For example, referring to fig. 2a, assuming that the sampled start node includes a node y, in the directed weighted graph shown in fig. 2a, the weighted walk starts from the node y, and when the weighted walk is based on the weight walk of each first directed edge connected by y, it is understood that the higher the weight of the first directed edge is, the greater the probability of selecting the first directed edge when the first directed edge is in the walk is, but the first directed edge with the largest weight is not necessarily selected. Assuming that the sampled node after the y node is x1, the walk will continue based on the weight of each first directed edge of the x1 connection, selecting the next node of x1, and so on until the number of geographic interest points in the geographic interest point visit sequence is equal to L.

104. Obtaining description vectors of geographic interest points;

in this embodiment, the description vector of the geographic interest point may be any description vector related to the geographic interest point in the prior art, and the description vector may be obtained based on any information of the geographic interest point, such as attribute information of the geographic interest point, user experience information, and the like. The attribute information of the geographic interest point includes, but is not limited to, information on attributes such as a name, an address, coordinates, a category, a social function, and the like. User experience information includes, but is not limited to, user comment information, and the like.

Optionally, in consideration of the fact that mining depth of check-in data of a user is not enough in the prior art, the embodiment performs semantic feature extraction on a sentence level on check-in texts of the POIs by acquiring the description vectors, and deeply mines rich text information in POI comments, so that effectiveness of knowledge representation of the POIs can be improved.

Optionally, the step of "obtaining a description vector of a geographic interest point" may include:

acquiring a user check-in text of the geographic interest points from historical check-in data;

semantic analysis is carried out on the user check-in text of the geographic interest points to obtain text description vectors for describing the semantics of the user check-in text, wherein the text description vectors are the same in length;

In this embodiment, the user check-in text may include any text content related to the corresponding geographic point of interest in the historical check-in data, including but not limited to types of text information such as a category of the POI, a feature of the POI, and a user experience. The categories of the geographic interest points may include categories of geographic entities, such as schools, banks, restaurants, gas stations, hospitals, supermarkets, scenic spots, shopping malls, and the like, and the features of the geographic interest points may include features of the geographic entities of each category, for example, for a shopping mall, the features may include: the size of the mall, the amount of people, the type of the transaction item, and the like, and the user experience information of the geographic interest point may include user comment information of the user on the geographic interest point, and the like. The user comment information can comprise comment texts input by the user, and can also comprise comment scores input by the user in a comment page of the geographic interest points, and the like.

In this embodiment, the step of obtaining the user check-in text of the geographic point of interest from the historical check-in data may include:

acquiring all check-in data corresponding to all geographic interest points from historical check-in data;

and acquiring preset text data from the check-in data of each geographic interest point to serve as the check-in text of the user of the geographic interest point.

The preset type may include the types of text information such as the categories of the aforementioned POIs, the features of the POIs, and the user experience.

In this embodiment, the number of user check-in texts of the geographic interest point is related to the number of users checking in at the geographic interest point, and one check-in of each user for the geographic interest point may generate one user check-in text for the geographic interest point.

In this embodiment, for each geographic interest point, since a plurality of users generally check in at the geographic interest point, the number of user check-in texts corresponding to the geographic interest point is generally multiple, and the semantic information of the geographic interest point can be obtained by mining the multiple user check-in texts corresponding to each geographic interest point.

In this embodiment, the check-in text of the user may be analyzed through a semantic analysis model, which may be any model that can be used for semantic analysis of the text, for example, the semantic analysis model may be a dialog (BERT) model.

The step of performing semantic analysis on the user check-in text of the geographic interest point to obtain a text description vector for describing the semantics of the user check-in text may include:

and performing semantic analysis on the texts by adopting a preset BERT model for all the user check-in texts of each geographical interest point to obtain an embedded vector of each user check-in text. The embedded vector can be understood as an embedded representation of the semantics of the user check-in text, i.e. a text description vector for describing the semantics of the user check-in text.

For example, referring to FIG. 2b, assume a geographic point of interest (POI)_i(abbreviation P_i) There is a correspondence of n_iSlip user check-in text

Extracting an embedded representation e of each user check-in text using a BERT dialog model_k＝BERT(s_k),k＝1,2,…,n_iFinally, n is obtained_iVectors of the same length.

Optionally, in this embodiment, the manner of fusing all the text description vectors of the same geographic interest point includes, but is not limited to, averaging, weighted averaging, and the like.

Optionally, the step of "fusing all text description vectors of the same geographic interest point to obtain a fused vector, and using the fused backward vector as a description vector of a corresponding geographic interest point" may include: and averaging all the text description vectors of the same geographic interest point to obtain an average vector as the description vector of the corresponding geographic interest point.

Still referring to FIG. 2b, with the aforementioned geographic point of interest P_iFor example, for a geographic point of interest P_iN of (A) to (B)_iA text description vector e_k＝BERT(s_k),k＝1,2,…,n_iThese vectors can be averagedAs the geographic point of interest P_iSemantic embedded representation (i.e. description vector)

Specifically, the description vector can be obtained by averaging the description vectors in a depth-wise-average pooling (depth-wise averaging) manner

105. Learning the access rule of the geographic interest points in the geographic interest point access sequence based on the description vector of the geographic interest points to obtain access relation vectors corresponding to the geographic interest points, wherein the access relation vectors are used for representing the association of the corresponding geographic interest points and other geographic interest points on user access behaviors.

In this embodiment, the association between the geographic interest point (denoted as the first geographic interest point) and other geographic interest points in the user access behavior may be understood as n geographic interest points before the first geographic interest point and n geographic interest points after the first geographic interest point and possibly accessed by the user, and the user access probabilities of these geographic interest points.

Optionally, in this embodiment, the step "learning an access rule of the geographic interest point in the geographic interest point access sequence based on the description vector of the geographic interest point to obtain an access relationship vector corresponding to the geographic interest point" may include:

and learning the access rule of the geographic interest points in the geographic interest point access sequence through an access rule analysis model based on the description vector of the geographic interest points to obtain an access relation vector corresponding to the geographic interest points.

Optionally, the step of learning the access rule of the geographic interest point in the geographic interest point access sequence based on the description vector of the geographic interest point through the access rule analysis model to obtain an access relationship vector corresponding to the geographic interest point may include:

learning user access rules of associated interest points corresponding to the geographic interest points in the geographic interest point access sequence through an access rule analysis model based on the description vectors of the geographic interest points, wherein the associated interest points corresponding to the geographic interest points are as follows: geographic interest points visited by the user before and after the geographic interest points in the geographic interest point visiting sequence;

determining an access relation vector of each geographic interest point based on a learning result of an access rule analysis model, wherein the access relation vector comprises candidate geographic interest points which are possibly accessed by a user before and after the corresponding geographic interest point and user access probability of the candidate geographic interest points.

In this embodiment, the user access rules include, but are not limited to, a relationship between the associated interest points corresponding to the geographic interest points and the geographic interest points in the user access sequence, and user access probabilities corresponding to the associated interest points corresponding to the geographic interest points.

Optionally, in this embodiment, the step "learning, by using an access rule analysis model, a user access rule of an associated interest point corresponding to each geographic interest point in the geographic interest point access sequence based on a description vector of the geographic interest point" may include:

arranging the description vectors of the geographic interest points according to the arrangement sequence of the geographic interest points in the geographic interest point access sequence to obtain a description vector sequence corresponding to the geographic interest points;

and taking the description vector sequence as a sentence, and learning the context of each description vector in the sentence through an access rule analysis model so as to learn the user access rule of the associated interest point corresponding to the geographic interest point in the geographic interest point access sequence.

In this embodiment, the access law analysis model may be any neural network model that can learn and output context information of each word in a sentence, for example, the access law analysis model may be a skip-gram model. The access relation vector corresponding to the geographic interest point can be a word vector corresponding to each description vector obtained by the skip-gram model based on the access sequence of the geographic interest point.

Taking a skip-gram model as an example, the process of acquiring the access relation vector is explained as follows:

taking the description vector sequence as a sentence, and performing context learning of each description vector through a skip-gram model;

and extracting an embedded vector corresponding to each description vector from a hidden layer of the skip-gram model as an access relation vector of a geographic interest point corresponding to the description vector.

In this embodiment, the embedded vector of the description vector describes context information of the description vector, such as occurrence probabilities of the description vectors at positions before and after the description vector, that is, occurrence rules of geographic interest points visited by the user before and after the geographic interest point corresponding to the description vector can be obtained by using the embedded vector.

The association of the geographic interest point with other geographic interest points in the user access behavior may include the geographic interest points that may appear in n access positions before and after the geographic interest point, and the user access probabilities of the geographic interest points.

Optionally, for the access law analysis model, in order to improve the accuracy of the access relationship vector, the access law analysis model may be pre-trained, and the pre-training of the access law analysis model may be performed based on a one-hot (one-hot) code of the geographic interest point.

Optionally, in this embodiment, before the step "learning the access rule of the geographic interest point in the geographic interest point access sequence based on the description vector of the geographic interest point through the access rule analysis model to obtain the access relationship vector corresponding to the geographic interest point", the method may further include:

and training the access law analysis model through the unique hot code sequence so that the access law analysis model learns the access law of the geographic interest points in the geographic interest point access sequence.

In this embodiment, when the access law analysis model is trained through the unique hot code sequence, it may be determined whether the access law analysis model is trained based on training times, training duration, or a loss function of the model, for example, the step "training the access law analysis model through the unique hot code sequence so that the access law analysis model learns the access law of the geographic interest point in the geographic interest point access sequence" may include:

training the access rule analysis model through the unique hot coding sequence so that the access rule analysis model learns the user access rules of the associated interest points corresponding to the geographic interest points in the geographic interest point access sequence;

and stopping training the access law analysis model when the training times of the access law analysis model reach a preset time threshold value.

In this embodiment, the access rule analysis model and the semantic analysis model may be jointly trained, and a description vector output by the semantic analysis model is used as a vector corresponding to a geographic interest point in the geographic interest point access sequence input to the semantic analysis model.

In this embodiment, in the process of training the access rule analysis model based on the one-hot code, before the access rule analysis model and the semantic analysis model can be jointly trained, parameters of the semantic analysis model are kept unchanged in the process of training the access rule analysis model based on the one-hot code.

For example, referring to FIG. 2b, when the skip-gram model is trained based on the one-hot of the POI, the parameters of the dialog model remain fixed.

In this embodiment, for each geographic interest point, the information of the geographic interest points which are possibly visited by the user at the front n visit positions and the back n visit positions of the geographic interest point can be obtained through the skip-gram model.

For example, referring to FIG. 2b, for a geographic point of interest POI_iThe dialogue model analyzes the sign-in text of the user to finally obtain the sign-in textDescription vector E^smtThe description vector E^smtInputting a kip-gram model, which can be analyzed to obtain POI_iThe access relationship vector may include POIs_iCandidate geographical points of interest (POI) at previous n visiting locations_i-n、POI_i-(n-1)、POI_i-(n-2)…POI_i-1And POI_iCandidate geographical points of interest (POI) on the next n visiting positions_i+1、POI_i+2、POI_i+3…POI_i+nThe information of (1). Optionally, in this embodiment, for the access relationship vector, POI_iOne or more candidate geographic interest points can be predicted in each of the previous and subsequent n visiting locations, and the number of predicted geographic interest points is not limited in this embodiment.

In this embodiment, the step "learning an access rule of the geographic interest point in the geographic interest point access sequence based on the description vector of the geographic interest point to obtain an access relationship vector corresponding to the geographic interest point" may further include: and fusing the description vector corresponding to the geographic interest point and the access relation vector to obtain a comprehensive description vector of the geographic interest point.

Optionally, the step of fusing the description vector corresponding to the geographic interest point with the access relationship vector to obtain a comprehensive description vector of the geographic interest point may include: and splicing the description vector corresponding to the geographic interest point and the access relation vector to obtain a comprehensive description vector of the geographic interest point.

For example, a description vector E for a geographic point of interest^smtAnd an access relation vector E^poiAnd splicing the two vectors to obtain a comprehensive description vector E ═ E (E) of the geographic interest point^smt,E^poi)。

Alternatively, for E^smtAnd E^poiCan be implemented by a skip-gram model, e.g. referring to FIG. 2b, the kip-gram model pair E^smtAnd E^poiAnd (6) splicing.

106. And generating related recommendation information of the geographic interest points based on the access relation vector.

Optionally, the step "generating relevant recommendation information of the geographic point of interest based on the access relationship vector" may include: when a recommendation request aiming at a target geographic interest point is received, an access relation vector corresponding to the target geographic interest point is obtained, a recommendation geographic interest point meeting the recommendation purpose of the recommendation request is determined based on the access relation vector, and relevant recommendation information responding to the recommendation request is generated based on the recommendation geographic interest point.

In this embodiment, the recommendation request may be sent by a terminal, and optionally, the step "obtaining an access relationship vector corresponding to a target geographic interest point when the recommendation request for the target geographic interest point is received" may include: when a recommendation request sent by a terminal is received, determining a geographic interest point where the terminal is located at present as a target geographic interest point, and acquiring an access relation vector corresponding to the target geographic interest point.

The recommendation request can carry information of the geographic interest point where the terminal is currently located, and the geographic interest point where the terminal is currently located can be determined as the target geographic interest point through the information of the geographic interest point where the terminal is currently located carried in the recommendation request.

Optionally, the recommendation request for the target geographic interest point in the embodiment may be for requesting to recommend the geographic interest point, or may be for requesting to recommend the route, which is not limited in this embodiment.

Optionally, in an embodiment, the recommendation request is a location recommendation request for a target geographic interest point, and the generated relevant recommendation information is a recommended geographic interest point.

Optionally, the step of determining, based on the access relationship vector, recommended geographic interest points that satisfy the recommendation purpose of the recommendation request, and generating relevant recommendation information in response to the recommendation request based on the recommended geographic interest points may include:

selecting recommended geographic interest points from the candidate geographic interest points based on the user access probability of the candidate geographic interest points in the access relation vector of the target geographic interest points;

and generating recommendation information responding to the position recommendation request based on the recommended geographic interest points, wherein the recommendation information comprises the position information of the recommended geographic interest points.

Referring to FIG. 2b, suppose that the geographic point of interest of user A is POI_iWhen the user A needs the server to recommend the geographical interest point which is accessed next, the user A sends a position recommendation request to the server through operation on the terminal. After the server receives the position recommendation request, the current geographic interest point of the user A is determined to be the POI_iTo POI_iObtaining POI as target geographical interest point_iA corresponding access relation vector.

Referring to FIG. 2b, POI_iThe access relationship vector including POI_iCandidate geographical points of interest (POI) at previous n visiting locations_i-n、POI_i-(n-1)、POI_i-(n-2)…POI_i-1And POI_iCandidate geographical points of interest (POI) on the next n visiting positions_i+1、POI_i+2、POI_i+3…POI_i+nAnd the user access probability of each candidate geographic interest point.

In determining a target geographic point of interest as a POI_iThereafter, the server may retrieve the POI from the POI_iThe candidate geographic interest point with the maximum user visit probability is selected as the recommended geographic interest point in the visit relation vector. Or, from POI_iIn the access relationship vector, the candidate geographic interest point with the highest user access probability is selected as the recommended geographic interest point, such as from the POI_i-1And POI_i+1And selecting the user with the highest access probability as the recommended geographic interest point. And then, acquiring relevant information (including but not limited to position information) of the recommended geographic interest points to generate recommendation information, and feeding the recommendation information back to the terminal, so that the terminal can display the recommended geographic interest points to the user A.

Optionally, in another embodiment, the recommendation request is a path recommendation request for the target geographic interest point; and the generated related recommendation information is a recommendation path.

The step of determining a recommended geographic point of interest that satisfies a recommendation purpose of the recommendation request based on the access relationship vector and generating relevant recommendation information in response to the recommendation request based on the recommended geographic point of interest may include:

determining recommended geographic interest points on a recommended path to be generated and a user access sequence of the recommended geographic interest points based on the user access probability of each candidate geographic interest point in the access relation vector of the target geographic interest points;

and generating a recommendation path based on the user access sequence of the recommended geographic interest points, wherein the recommendation path is the relevant recommendation information responding to the path recommendation request.

Still referring to FIG. 2b, assume that the geographic point of interest at which user A is currently located is a POI_iWhen the user A needs the server to recommend the travel route, the user A sends a route recommendation request to the server through operation on the terminal. After the server receives the path recommendation request, determining that the current geographic interest point of the user A is the POI_iTo POI_iObtaining POI as target geographical interest point_iA corresponding access relation vector.

The server determines the target geographical interest point as the POI_iThen, a candidate geographic interest point can be respectively selected from a plurality of visiting positions behind the visiting position of the target geographic interest point to serve as a recommended geographic interest point of the visiting position, and a recommended path is generated by taking the sequence of the visiting positions as the visiting sequence of the recommended geographic interest points. E.g. from POI_i+1、POI_i+2、POI_i+3…POI_i+nThe n access positionsAnd respectively selecting a candidate geographic interest point with the maximum user access probability as a recommended geographic interest point at the first m access positions, and generating a recommended path by using the recommended geographic interest points at the m access positions. And then sending the recommended path to the terminal of the user A so that the terminal displays the recommended path to the user A.

The category of the geographic interest point can be manually labeled, but the manual labeling of the category of the geographic interest point has high demand on human resources.

Optionally, the method of this embodiment further includes:

acquiring a training sample, wherein the training sample is a user check-in text of a sample geographical interest point of a user historical check-in, and a label of the training sample comprises an expected category of the sample geographical interest point corresponding to the user check-in text;

and training a text analysis model by using the training samples.

In this embodiment, the geographic areas of the obtained sample geographic interest points and the geographic interest points in step 101 may be the same, but the obtaining time may be different.

The training of the text analysis model with the training samples is to learn the association relationship between the training samples and the expected classes in the labels for the text analysis model. Optionally, the step of "training a text analysis model with the training sample" may include: analyzing the user sign-in text of the sample geographical interest points in the training samples through a text analysis model, and predicting the categories of the sample geographical interest points; parameters of the text analysis model are adjusted based on the predicted and expected categories of the sample geographic points of interest.

In this embodiment, the text analysis model may be a classification model that can be used for text analysis in the related art.

Optionally, before the step of performing semantic analysis on the user check-in text of the geographic interest point to obtain a text description vector for describing semantics of the user check-in text, "the method may further include:

analyzing the user check-in text of the geographical interest points through the text analysis model to obtain the prediction categories of the geographical interest points;

and adding the prediction categories of the geographic interest points into user check-in texts corresponding to the geographic interest points.

By the method, the category information of the geographic interest points can be automatically added into the check-in text of the user.

Optionally, in view of resource waste caused by repeatedly adding category information, the step "analyzing the user check-in text of the geographic interest point through the text analysis model to obtain the prediction category of the geographic interest point" in this embodiment may include: and analyzing the user check-in text of the geographical interest points through the text analysis model to obtain the predicted category of the geographical interest points, wherein the geographical interest points do not contain the category information of the geographical interest points in the user check-in text.

In this embodiment, the comprehensive description vector of the geographic interest point includes a description vector for describing a semantic meaning of a check-in text of the user, so it can be understood that the category of the geographic interest point can be determined based on the comprehensive description vector of the geographic interest point.

In one example, the server may predict a next geographical interest point type preferred by the user using the terminal when receiving the location recommendation request sent by the terminal, and select the recommended geographical interest point based on the predicted type and the access relationship vector of the geographical interest point where the terminal is located.

determining a target type of recommended geographic interest points based on access relation information among the geographic interest points, wherein the probability that the geographic interest points under the target type are accessed by a user after the target geographic interest points is the largest;

obtaining an access relation vector of a target geographic interest point, and determining type information and user access probability of a candidate geographic interest point of the target geographic interest point based on the access relation vector;

selecting the geographical interest points which belong to the target type and have the maximum user access probability from the candidate geographical interest points as recommended geographical interest points;

The server can determine POI based on the access relation information between the geographic interest points_iThe target type of the subsequent geographic point of interest. For example, the server discovers, at the POI, based on the access relationship information of the geographic points of interest_iThe geographic points of interest visited later are 1/2 of mall type, 1/4 of park type and 1/4 of restaurant type. The type of the target for which the mall is a geographic point of interest is determined.

In determining a target geographic point of interest as a POI_iThereafter, the server may retrieve the POI from the POI_iThe candidate geographic interest point which belongs to the shopping mall type and has the highest user access probability is selected as the recommended geographic interest point in the access relation vector. Or, based on POI_iAccess relationship ofVector for selecting the candidate geographical interest point with the highest user visit probability as the recommended geographical interest point from the candidate geographical interest points with the visit positions closest to the visit position of the target geographical interest point and belonging to the shopping mall type, such as POI_i-1And POI_i+1And selecting the user with the highest access probability which belongs to the shopping mall type as the recommended geographic interest point. And then, acquiring relevant information (including but not limited to position information) of the recommended geographic interest points to generate recommendation information, and feeding the recommendation information back to the terminal, so that the terminal can display the recommended geographic interest points to the user A.

The embodiment discloses a recommendation method based on interest points, which can acquire historical check-in data of a user and determine geographical interest points of the historical check-in of the user from the historical check-in data; acquiring access relation information between geographic interest points based on historical sign-in data; generating a geographical interest point access sequence based on access relation information among the geographical interest points, wherein adjacent geographical interest points in the geographical interest point access sequence are accessed by at least one user successively; obtaining description vectors of geographic interest points; learning the access rules of the geographic interest points in the geographic interest point access sequence through a preset access rule analysis model based on the description vectors of the geographic interest points to obtain access relation vectors corresponding to the geographic interest points, wherein the access relation vectors are used for representing the association of the corresponding geographic interest points and other geographic interest points on user access behaviors, the distinctiveness of the geographic interest points on user dimensions is weakened in the embodiment, the check-in sequence of each user is not taken as a research object, but is disassembled, the geographic interest point access sequence is constructed based on the access relations among the geographic interest points, the sparsity of the user check-in sequence is compensated, the representation of the association of the geographic interest points on the user access behaviors is promoted, and when the corresponding geographic interest points are required to be recommended for the target geographic interest points, the point of interest which is relatively large in relation to the geographic point of interest in the user access behavior can be determined based on the access relation vector, and therefore accuracy and reliability of recommendation services based on the relation are improved.

In order to better implement the above method, correspondingly, an embodiment of the present invention further provides a point of interest-based recommendation apparatus, which may be integrated in a terminal, and the point of interest-based recommendation apparatus includes, with reference to fig. 3:

the check-in data acquisition unit 301 is configured to acquire historical check-in data of a user, and determine geographic interest points of historical check-in of the user from the historical check-in data;

the access relation obtaining unit 302 is configured to obtain access relation information between geographic interest points based on historical check-in data;

the sequence generating unit 303 is configured to generate a geographic interest point access sequence based on access relationship information between geographic interest points, where adjacent geographic interest points in the geographic interest point access sequence are accessed by at least one user in sequence;

a description vector obtaining unit 304, configured to obtain a description vector of the geographic interest point;

an access relationship vector obtaining unit 305, configured to learn an access rule of a geographic interest point in a geographic interest point access sequence through a preset access rule analysis model based on a description vector of the geographic interest point, to obtain an access relationship vector corresponding to the geographic interest point, where the access relationship vector is used to represent an association between the corresponding geographic interest point and other geographic interest points in a user access behavior;

and the recommending unit 306 is configured to generate relevant recommendation information of the geographic interest point based on the access relationship vector of the geographic interest point.

Optionally, the description vector obtaining unit includes:

Optionally, the access relationship obtaining unit includes:

Optionally, the sequence generating unit includes:

Optionally, the user preference degree includes a user access number;

a generating subunit for: generating nodes of a directed weighted graph based on the geographic interest points, and generating directed edges among the nodes based on the user access directions among the geographic interest points; calculating the user access times and the total user access times corresponding to a first directed edge pointing to other geographic interest points from the same geographic interest point in the directed weighted graph; and calculating the ratio of the user access times of the connected first directed edges to the total corresponding user access times of the same geographic interest point, and taking the ratio as the weight of the corresponding first directed edge.

Optionally, the apparatus of this embodiment further includes: and the fusion unit is used for learning the access rule of the geographic interest points in the geographic interest point access sequence through a preset access rule analysis model on the basis of the description vectors of the geographic interest points in the access relation vector acquisition unit to obtain access relation vectors corresponding to the geographic interest points, and then fusing the description vectors corresponding to the geographic interest points with the access relation vectors to obtain comprehensive description vectors of the geographic interest points.

The embodiment discloses a recommendation device based on interest points, which does not take a check-in sequence of each user as a research object, but disassembles the check-in sequence of each user, constructs a geographic interest point access sequence based on an access relation between geographic interest nodes, makes up for sparsity of the check-in sequence of the users, is favorable for improving representation of association between the geographic interest points on user access behaviors, further, combines text information such as social comment information in historical check-in data of the users, deeply excavates rich text information in the comments, is favorable for improving effectiveness of knowledge representation of POI, and improves accuracy and reliability of related location service based on the POI,

in addition, an embodiment of the present invention further provides a computer device, where the computer device may be a terminal or a server, as shown in fig. 4, which shows a schematic structural diagram of the computer device according to the embodiment of the present invention, and specifically:

the computer device may include components such as a processor 401 of one or more processing cores, memory 402 of one or more computer-readable storage media, and a power supply 403. Those skilled in the art will appreciate that the computer device configuration illustrated in FIG. 4 does not constitute a limitation of computer devices, and may include more or fewer components than those illustrated, or some components may be combined, or a different arrangement of components. Wherein:

the processor 401 is a control center of the computer device, connects various parts of the entire computer device using various interfaces and lines, and performs various functions of the computer device and processes data by operating or executing software programs and/or units stored in the memory 402 and calling data stored in the memory 402, thereby monitoring the computer device as a whole. Optionally, in one embodiment, processor 401 may include one or more processing cores; preferably, the processor 401 may integrate an application processor, which mainly handles operating systems, user interfaces, application programs, etc., and a modem processor, which mainly handles wireless communications. It will be appreciated that the modem processor described above may not be integrated into the processor 401.

The memory 402 may be used to store software programs and units, and the processor 401 executes various functional applications and data processing by operating the software programs and units stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application program required by at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may store data created according to use of the computer device, and the like. Further, the memory 402 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid state storage device. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 access to the memory 402.

The computer device further comprises a power supply 403 for supplying power to the various components, and preferably, the power supply 403 is logically connected to the processor 401 via a power management system, so that functions of managing charging, discharging, and power consumption are implemented via the power management system. The power supply 403 may also include any component of one or more dc or ac power sources, recharging systems, power failure detection circuitry, power converters or inverters, power status indicators, and the like.

When the computer device is a terminal, the computer device may further include an input unit 404, and the input unit 404 may be used to receive input numeric or character information and generate a keyboard, mouse, joystick, optical or trackball signal input in relation to user setting and function control. Of course, it is understood that the present embodiment does not exclude the solution that the server includes the input unit, and the server of the present embodiment may also include the input unit 404.

Although not shown, the computer device, such as the terminal, of the present embodiment may further include a display unit and the like, which are not described herein again. Similarly, the present embodiment does not exclude the scheme that the server includes the display unit, and the server in the present embodiment may also include the display unit.

Specifically, in this embodiment, the processor 401 in the computer device loads the executable file corresponding to the process of one or more application programs into the memory 402 according to the following instructions, and the processor 401 runs the application programs stored in the memory 402, thereby implementing various functions as follows:

acquiring historical sign-in data of a user, and determining geographic interest points of the historical sign-in of the user from the historical sign-in data;

acquiring access relation information between geographic interest points based on historical sign-in data;

generating a geographical interest point access sequence based on access relation information among the geographical interest points, wherein adjacent geographical interest points in the geographical interest point access sequence are accessed by at least one user successively;

obtaining description vectors of geographic interest points;

learning the access rule of the geographic interest points in the geographic interest point access sequence through an access rule analysis model based on the description vector of the geographic interest points to obtain access relation vectors corresponding to the geographic interest points, wherein the access relation vectors are used for expressing the association of the corresponding geographic interest points and other geographic interest points on user access behaviors;

The above operations can be implemented in the foregoing embodiments, and are not described in detail herein.

Therefore, the computer equipment of the embodiment can deeply mine rich text information in POI comments, can improve the effectiveness of knowledge representation of the POI, can avoid the defect of sparse data existing when the check-in sequence of each user is used as a research object, and establishes the association between the POI. And further, accuracy and reliability of related location services, such as POI classification prediction, route planning, location recommendation and other users, are improved.

The point-of-interest-based recommendation system according to the embodiment of the present invention may be a distributed system formed by connecting a client, a plurality of nodes (any form of computer devices in an access network, such as servers and terminals) through network communication.

Taking a distributed system as an example of a blockchain system, referring to fig. 5, fig. 5 is an optional structural schematic diagram of the distributed system 100 applied to the blockchain system, which is formed by a plurality of nodes (computing devices in any form in an access network, such as servers and user terminals) and clients, and a Peer-to-Peer (P2P, Peer to Peer) network is formed between the nodes, and the P2P Protocol is an application layer Protocol operating on a Transmission Control Protocol (TCP). In a distributed system, any machine, such as a server or a terminal, can join to become a node, and the node comprises a hardware layer, a middle layer, an operating system layer and an application layer. In this embodiment, historical check-in data, an access rule analysis model, a semantic analysis model, a text analysis model, a directed weighted graph, a description vector of a geographic interest point, an access relation vector of the geographic interest point, a comprehensive description vector of the geographic interest point, and the like may all be stored in a shared ledger of a regional chain system through nodes of a distributed system, a computer device (e.g., a terminal or a server) may obtain historical check-in data based on recorded data stored in the shared ledger, and the access rule analysis model, the semantic analysis model, the comprehensive description vector, and the like.

Referring to the functions of each node in the blockchain system shown in fig. 5, the functions involved include:

1) routing, a basic function that a node has, is used to support communication between nodes.

Besides the routing function, the node may also have the following functions:

2) the application is used for being deployed in a block chain, realizing specific services according to actual service requirements, recording data related to the realization functions to form recording data, carrying a digital signature in the recording data to represent a source of task data, and sending the recording data to other nodes in the block chain system, so that the other nodes add the recording data to a temporary block when the source and integrity of the recording data are verified successfully.

For example, the services implemented by the application include:

2.1) wallet, for providing the function of transaction of electronic money, including initiating transaction (i.e. sending the transaction record of current transaction to other nodes in the blockchain system, after the other nodes are successfully verified, storing the record data of transaction in the temporary blocks of the blockchain as the response of confirming the transaction is valid; of course, the wallet also supports the querying of the remaining electronic money in the electronic money address;

and 2.2) sharing the account book, wherein the shared account book is used for providing functions of operations such as storage, query and modification of account data, record data of the operations on the account data are sent to other nodes in the block chain system, and after the other nodes verify the validity, the record data are stored in a temporary block as a response for acknowledging that the account data are valid, and confirmation can be sent to the node initiating the operations.

2.3) Intelligent contracts, computerized agreements, which can enforce the terms of a contract, implemented by codes deployed on a shared ledger for execution when certain conditions are met, for completing automated transactions according to actual business requirement codes, such as querying the logistics status of goods purchased by a buyer, transferring the buyer's electronic money to the merchant's address after the buyer signs for the goods; of course, smart contracts are not limited to executing contracts for trading, but may also execute contracts that process received information.

3) And the Block chain comprises a series of blocks (blocks) which are mutually connected according to the generated chronological order, new blocks cannot be removed once being added into the Block chain, and recorded data submitted by nodes in the Block chain system are recorded in the blocks.

Referring to fig. 6, fig. 6 is an optional schematic diagram of a Block Structure (Block Structure) according to an embodiment of the present invention, where each Block includes a hash value of a transaction record stored in the Block (hash value of the Block) and a hash value of a previous Block, and the blocks are connected by the hash values to form a Block chain. The block may include information such as a time stamp at the time of block generation. A block chain (Blockchain), which is essentially a decentralized database, is a string of data blocks associated by using cryptography, and each data block contains related information for verifying the validity (anti-counterfeiting) of the information and generating a next block.

It will be understood by those skilled in the art that all or part of the steps of the methods of the above embodiments may be performed by instructions or by associated hardware controlled by the instructions, which may be stored in a computer readable storage medium and loaded and executed by a processor.

To this end, the present invention further provides a storage medium, in which a plurality of instructions are stored, and the instructions can be loaded by a processor to execute the steps in any of the point of interest-based recommendation methods provided by the embodiments of the present invention.

Wherein the storage medium may include: read Only Memory (ROM), Random Access Memory (RAM), magnetic or optical disks, and the like.

Since the instructions stored in the storage medium may execute the steps in any point-of-interest-based recommendation method provided in the embodiments of the present invention, beneficial effects that can be achieved by any point-of-interest-based recommendation method provided in the embodiments of the present invention may be achieved, for details, see the foregoing embodiments, and are not described herein again.

The recommendation method, apparatus, computer device and storage medium based on the point of interest provided by the embodiment of the present invention are introduced in detail above, and a specific example is applied in the present document to explain the principle and the implementation of the present invention, and the description of the above embodiment is only used to help understanding the method and the core idea of the present invention; meanwhile, for those skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.

Claims

1. A point of interest-based recommendation method is characterized by comprising the following steps:

obtaining a description vector of the geographic interest point;

2. The method of claim 1, wherein the obtaining a description vector of the geographic point of interest comprises:

3. The interest point-based recommendation method according to claim 2, wherein the fusing all the text description vectors of the same geographic interest point to obtain a fused vector, and using the fused backward vector as the description vector of the corresponding geographic interest point comprises:

4. The interest point-based recommendation method according to claim 2, wherein the learning of the access rules of the geographic interest points in the geographic interest point access sequence based on the description vector of the geographic interest points to obtain the access relationship vector corresponding to the geographic interest points comprises:

before learning the access rule of the geographic interest point in the geographic interest point access sequence based on the description vector of the geographic interest point through the access rule analysis model to obtain the access relationship vector corresponding to the geographic interest point, the method further comprises the following steps:

5. The point-of-interest-based recommendation method according to claim 2, wherein said obtaining access relationship information between said geographic points of interest based on said historical check-in data comprises:

generating a geographic interest point access sequence based on the access relation information among the geographic interest points, wherein the generating comprises the following steps:

6. The point-of-interest recommendation method according to claim 5, wherein the generating a geographic point-of-interest visit sequence of a preset length based on the weights of the nodes and the directed edges in the directed weighted graph comprises:

7. The point-of-interest-based recommendation method according to claim 5, wherein said user preference degree comprises user access times;

8. The interest point-based recommendation method according to any one of claims 1-7, wherein the method for learning the access rules of the geographic interest points in the geographic interest point access sequence based on the description vector of the geographic interest points further comprises:

9. A point of interest based recommendation apparatus, comprising:

10. A storage medium having a computer program stored thereon, wherein the computer program when executed by a processor implements the steps of the method according to any of claims 1-8.

11. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the steps of the method according to any of claims 1-8 are implemented when the program is executed by the processor.