CN113449210A - Personalized recommendation method and device based on space-time characteristics, electronic equipment and storage medium - Google Patents
Personalized recommendation method and device based on space-time characteristics, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN113449210A CN113449210A CN202110747365.0A CN202110747365A CN113449210A CN 113449210 A CN113449210 A CN 113449210A CN 202110747365 A CN202110747365 A CN 202110747365A CN 113449210 A CN113449210 A CN 113449210A
- Authority
- CN
- China
- Prior art keywords
- information
- matrix
- implicit
- browsing
- browsed
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9537—Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/958—Organisation or management of web site content, e.g. publishing, maintaining pages or automatic linking
Abstract
The embodiment of the invention provides a personalized recommendation method based on space-time characteristics, which comprises the following steps: acquiring a time-space matrix of a target user when browsing information, wherein the time-space matrix comprises browsing time, browsing places and browsing labels, and different browsing labels correspond to different types of information contents; acquiring a scoring matrix of all information, wherein the scoring matrix comprises information tags, user information and scoring values, and different information tags correspond to different types of information contents; carrying out implicit factorization on the scoring matrix to obtain an implicit information matrix based on the information label, wherein the implicit information matrix comprises the information label and a first implicit score; obtaining an individualized recommendation matrix of a target user based on the implicit information matrix and the space-time matrix; and carrying out personalized recommendation on the target user according to the personalized recommendation matrix. The method and the device can consider the time and place where the user is interested in the information, so that the browsing personality of the user is fully mined, and the recommendation accuracy is improved.
Description
Technical Field
The invention relates to the field of intelligent recommendation, in particular to a personalized recommendation method and device based on space-time characteristics, electronic equipment and a storage medium.
Background
The information push is to recommend specific information to a user so as to reduce the search cost of the user for the information, and with the development of the internet and the development of big data, the information push can search information which may be interesting to the user according to the interest of the user, filter information which may not be interesting to the user, push the information which is interesting to the user, and help the user browse the information which is interesting to the user with high efficiency.
The existing recommendation method is mainly used for recommending through information similarity or user groups, and the recommendation based on the information similarity refers to that when a user is interested in a certain information content, the same type of information content is recommended for the user, and the recommendation based on the user groups refers to that the users are clustered, and the same type of information content is recommended for the users. However, both recommendation based on information similarity and recommendation based on user groups have the problem of solidification of recommendation information, and can recommend homogeneous information content to a user for a long time.
Disclosure of Invention
The embodiment of the invention provides a personalized recommendation method based on space-time characteristics, which can be used for recommending information to a user according to the information of the user in time and space on the basis of information scoring.
In a first aspect, an embodiment of the present invention provides a method for personalized recommendation based on spatio-temporal features, where the method includes:
acquiring a time-space matrix of a target user when browsing information, wherein the time-space matrix comprises browsing time, browsing places and browsing labels, and different browsing labels correspond to different types of information contents;
acquiring a scoring matrix of all information, wherein the scoring matrix comprises information tags, user information and scoring values, and different information tags correspond to different types of information contents;
carrying out implicit factorization on the scoring matrix to obtain an implicit information matrix based on an information label, wherein the implicit information matrix comprises the information label and a first implicit score;
obtaining an individualized recommendation matrix of the target user based on the implicit information matrix and the space-time matrix;
and carrying out personalized recommendation on the target user according to the personalized recommendation matrix.
Optionally, the obtaining of the scoring matrix of all information includes:
acquiring historical browsing information of a platform, wherein the historical browsing information comprises an information tag of browsed information, browsed times, positive evaluation times, negative evaluation times and user information corresponding to the browsed information;
calculating the score value of the browsed information according to the browsed times, the positive evaluation times and the negative evaluation times of the browsed information;
and constructing the scoring matrix according to the information tag of the browsed information, the scoring value of the browsed information and the user information corresponding to the browsed information.
Optionally, the obtaining of the personalized recommendation matrix of the target user based on the implicit information matrix and the spatio-temporal matrix includes:
masking the space-time matrix to obtain a mask matrix of the space-time matrix, wherein one mask value corresponds to one browsing label in the mask matrix;
and obtaining the personalized recommendation matrix of the target user based on the implicit information matrix and the mask matrix.
Optionally, before obtaining the personalized recommendation matrix of the target user based on the implicit information matrix and the mask matrix, obtaining the personalized recommendation matrix of the target user based on the implicit information matrix and the spatio-temporal matrix, further includes:
and carrying out dimension change on the mask matrix so that the dimension of the mask matrix is the same as that of the implicit information matrix.
Optionally, the performing dimensional change on the mask matrix includes:
and carrying out dimension change on the mask matrix through a preset full convolution network, wherein the output dimension of the full convolution network is the same according to the dimension of the implicit information matrix.
Optionally, the masking the space-time matrix to obtain a mask matrix of the space-time matrix includes:
carrying out implicit factorization on the scoring matrix to obtain an implicit user matrix based on user information, wherein the implicit user matrix comprises the user information and a second implicit score;
calculating a mask coefficient of the target user according to the second implicit score;
and performing mask calculation on the mask matrix according to the mask coefficient.
In a second aspect, an embodiment of the present invention provides a device for personalized recommendation based on spatiotemporal features, where the device includes:
the system comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring a spatiotemporal matrix of a target user when browsing information, the spatiotemporal matrix comprises browsing time, browsing place and browsing labels, and different browsing labels correspond to different types of information contents;
the second acquisition module is used for acquiring a scoring matrix of all information, wherein the scoring matrix comprises information tags, user information and scoring values, and different information tags correspond to different types of information contents;
the decomposition module is used for carrying out implicit factorization on the scoring matrix to obtain an implicit information matrix based on an information label, and the implicit information matrix comprises the information label and a first implicit score;
the processing module is used for obtaining the personalized recommendation matrix of the target user based on the implicit information matrix and the space-time matrix;
and the recommending module is used for carrying out personalized recommendation on the target user according to the personalized recommendation matrix.
Optionally, the second obtaining module includes:
the acquisition submodule is used for acquiring historical browsing information of the platform, wherein the historical browsing information comprises an information tag of browsed information, browsed times, positive evaluation times, negative evaluation times and user information corresponding to the browsed information;
the calculation submodule is used for calculating the score value of the browsed information according to the browsed times, the positive evaluation times and the negative evaluation times of the browsed information;
and the construction sub-module is used for constructing the scoring matrix according to the information tag of the browsed information, the scoring value of the browsed information and the user information corresponding to the browsed information.
In a third aspect, an embodiment of the present invention provides an electronic device, including: the system comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the computer program to realize the steps of the personalized recommendation method based on the spatiotemporal characteristics provided by the embodiment of the invention.
In a fourth aspect, the embodiment of the present invention provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the computer program implements the steps in the method for personalized recommendation based on spatiotemporal features provided by the embodiment of the present invention.
In the embodiment of the invention, a time-space matrix of a target user during information browsing is obtained, wherein the time-space matrix comprises browsing time, browsing place and browsing labels, and different browsing labels correspond to different types of information contents; acquiring a scoring matrix of all information, wherein the scoring matrix comprises information tags, user information and scoring values, and different information tags correspond to different types of information contents; carrying out implicit factorization on the scoring matrix to obtain an implicit information matrix based on an information label, wherein the implicit information matrix comprises the information label and a first implicit score; obtaining an individualized recommendation matrix of the target user based on the implicit information matrix and the space-time matrix; and carrying out personalized recommendation on the target user according to the personalized recommendation matrix. The method and the device can recommend the information to the user according to the information of the user in time and space on the basis of information scoring, and fully explores the browsing personality of the user due to the fact that the user is interested in the information when, where and so as to improve the recommendation accuracy.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a flowchart of a method for personalized recommendation based on spatiotemporal features according to an embodiment of the present invention;
fig. 2 is a flowchart of a score matrix acquisition according to an embodiment of the present invention;
FIG. 3 is a schematic structural diagram of a personalized recommendation device based on spatiotemporal features according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of a second obtaining module according to an embodiment of the present invention;
FIG. 5 is a schematic structural diagram of a processing module according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of another processing module according to an embodiment of the present invention;
fig. 7 is a schematic structural diagram of a mask submodule according to an embodiment of the present invention;
fig. 8 is a schematic structural diagram of an electronic device 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.
Referring to fig. 1, fig. 1 is a flowchart of a method for personalized recommendation based on spatiotemporal features according to an embodiment of the present invention, as shown in fig. 1, including the following steps:
101. and acquiring a space-time matrix of the target user when browsing information.
In an embodiment of the present invention, the spatio-temporal matrix includes browsing time, browsing location, and browsing tags, where different browsing tags correspond to different types of information content.
Further, the information browsed by the target user may be local platform information or cross-platform information, the browsing time refers to a time point when the target user browses the information, the browsing location refers to a location where the target user browses the information, and the browsing time and the browsing location may be obtained by a terminal device of the target user, for example, by a time built in the terminal device, when the target user browses the information, the corresponding browsing time is recorded, and when the target user browses the information, the corresponding browsing location is recorded by a positioning system built in the terminal device.
In the embodiment of the present invention, the platform is a platform on which the personalized recommendation method based on the spatio-temporal features in the embodiment of the present invention is mounted, the cross-platform refers to a platform including other platforms besides the platform, and the other platforms may not be platforms on which the personalized recommendation method based on the spatio-temporal features in the embodiment of the present invention is mounted. Under the condition of the platform, the browsing label which is allocated to the browsing information in advance can be directly acquired according to the browsing information of the target user. Under the cross-platform scenario, the corresponding browsing tag may be obtained according to the browsing information of the target user, for example, the browsing information of the target user is obtained, the browsing information of the target user is semantically identified, and then the browsing information of the target user is classified according to the semantic identification result, the classification standard is performed according to the classification standard of the platform, and then the browsing tag is marked on the browsing information of the target user according to the classification result. Thus, the same type of tag as the present platform classification standard can be obtained as the browsing tag.
In a possible embodiment, the browsing information in the spatio-temporal matrix may be non-recommended information and/or positive feedback information in recommended information, where the positive feedback information indicates that the target user gives a positive evaluation to the recommended information.
Specifically, the spatio-temporal matrix may be as shown in table 1 below:
TABLE 1
In table 1, Dl is the browsing time (in minutes or seconds), Tl is the browsing place, and l is the total number of times of browsing. It can be seen that the space-time matrix is a diagonal matrix, which facilitates arbitrary form transformation.
In one possible embodiment, the spatio-temporal matrix is a spatio-temporal matrix of the target user in a recent period, such as the spatio-temporal matrix of the user yesterday.
102. And acquiring a scoring matrix of all information.
In an embodiment of the present invention, the scoring matrix includes information tags, user information, and scoring values, where different information tags correspond to different types of information content. All of the above information includes all of the information of the present platform, or all of the above information includes all of the information of the cross platform.
The platform is a platform for carrying the space-time feature-based personalized recommendation method in the embodiment of the invention, the cross-platform means that the platform also comprises other platforms besides the platform, and the other platforms may not be platforms for carrying the space-time feature-based personalized recommendation method in the embodiment of the invention. Under the situation of the platform, the information labels pre-distributed to all information can be directly acquired. Under the cross-platform scenario, all information of each platform can be collected, collected browsed information is classified, and then corresponding information tags are obtained, for example, all browsed information of each platform is obtained, semantic recognition is performed on all browsed information, then all browsed information is classified according to the semantic recognition result, the classification standard is performed according to the classification standard of the platform, and then corresponding information tags are printed on all browsed information according to the classification result. Thus, the same type of tag as the present platform classification standard can be obtained as the information tag.
The browsing tab is directed to the target user, the information tab is directed to the information itself, and the browsing tab belongs to the information tab.
Specifically, the scoring matrix may be as shown in table 2 below:
TABLE 2
In table 1, User n is a User, mac m is a type of information, n is a total User number, m is a total information type number, and the value in the matrix unit is an average score of the corresponding User on the type of information.
Optionally, referring to fig. 2, fig. 2 is a flowchart of obtaining a scoring matrix according to an embodiment of the present invention, as shown in fig. 2, including the following steps:
201. and acquiring historical browsing information of the platform.
In an embodiment of the present invention, the historical browsing information includes an information tag of the browsed information, browsed times, positive evaluation times, negative evaluation times, and user information corresponding to the browsed information. The platform can be the current platform or a cross platform. The above-mentioned history browsing information refers to browsing information of each user.
The information label of the browsed information is a label obtained by classifying based on the platform; the browsed times refer to one type of information or the total browsed times of one information; the positive evaluation refers to that the user takes praise, praise and vote and leaves a message in a forward direction (such as the occurrence of words like, depth, good and the like) for one type of information or one piece of information, and the positive evaluation times refers to the times of the user taking praise, praise and vote, leaving a message in a forward direction and the like; the negative evaluation refers to that the user makes a click on, an objection to vote, a negative message (such as the occurrence of words like dislike, bad, and the like) for one type of information or one information, and the negative evaluation times refers to the times of making a click on, an objection to vote, a negative message, and the like.
202. And calculating the score value of the browsed information according to the browsed times, the positive evaluation times and the negative evaluation times of the browsed information.
In the embodiment of the present invention, the score value of the browsed information is calculated by a score of a certain type, and specifically, the score value of the browsed information may be calculated by the following equation:
in the above formula, Pi,jA value of a score of user i to information type j, ai,jFor the positive evaluation times of the user i for the information type j, bi,jFor the number of negative evaluations, V, of user i on information type ji,jThe information type j corresponds to the number of browsed times of the user i.
203. And constructing a grading matrix according to the information label of the browsed information, the grading value of the browsed information and the user information corresponding to the browsed information.
In the embodiment of the present invention, the scoring matrix shown in table 2 may be constructed by the information tag of the browsed information, the scoring value of the browsed information, and the user information corresponding to the browsed information.
103. And carrying out implicit factorization on the scoring matrix to obtain an implicit information matrix based on the information label.
In the embodiment of the present invention, the implicit information matrix includes an information tag and a first implicit score.
Specifically, the score matrix is subjected to implicit factorization, so that an implicit information matrix based on the information label and an implicit user matrix based on the user information can be obtained, and the implicit user matrix comprises the user information and a second implicit score.
Specifically, the implicit information matrix based on the information tag is shown in table 3 below:
TABLE 3
| U1 | U 2 | … | U k-1 | U k | |
| mac1 | u1,1 | u1,2 | … | u1,k-1 | u1,k |
| mac2 | u2,1 | u2,2 | … | u2,k-1 | u2,k |
| … | … | … | … | … | … |
| macm-1 | u m-1,1 | u m-1,2 | … | u m-1,k-1 | u m-1,k |
| macm | um,1 | um,2 | … | u1,k-1 | um,k |
In table 3 above, U is an implicit factor of the information tag.
The first time factor matrix is shown in table 4 below:
TABLE 4
| User1 | User2 | … | Usern-1 | Usern | |
| V1 | v1,1 | v1,1 | … | v1,n-1 | v1,n |
| V2 | v2,1 | v2,1 | … | v2,n-1 | v2,n |
| … | … | … | … | … | … |
| Vk-1 | v k-1,1 | v k-1,1 | … | v k-1,n-1 | Vk-1,n |
| Vk | vk,1 | vv,1 | … | vk,n-1 | vk,n |
In table 4 above, implicit factors for user information.
In the embodiment of the present invention, the scoring matrix in table 2 may be obtained by multiplying the implicit information matrix based on the information tag in table 3 and the implicit user matrix based on the user information in table 4. Specifically, the following formula can be used:
P＝UVT
in the formula, R is a scoring matrix, U is an implicit information matrix based on an information label, and V is an implicit user matrix based on user information.
104. And obtaining an individualized recommendation matrix of the target user based on the implicit information matrix and the space-time matrix.
In the embodiment of the invention, the implicit information matrix is based on the information label and is a matrix for carrying out implicit marking on the information label. The space-time matrix records information browsing conditions of a target user at different time and different positions in a near period of time, and the combination of the space-time matrix and the implicit information matrix is equivalent to weighting the first implicit score, so that the first implicit score can reflect information interested by the target user in the near period of time. In addition, because the implicit information matrix does not contain user information, the implicit information matrix can be better combined with the space-time matrix, and overfitting is avoided. The combination of the implicit information matrix and the spatio-temporal matrix may be a combination by matrix multiplication.
Furthermore, since the value in the matrix unit of the spatio-temporal matrix is a browsing label, not a numerical value, it is necessary to digitize the browsing label in the spatio-temporal matrix. The above-mentioned numeralization may be set according to the heat of each information type on the current platform or across platforms, and the higher the heat is, the higher the numerical value corresponding to the information type is.
In the embodiment of the invention, the space-time matrix can be masked to obtain a mask matrix of the space-time matrix, and one mask value corresponds to one browsing label in the mask matrix; and obtaining the personalized recommendation matrix of the target user based on the implicit information matrix and the mask matrix. The mask value is a value set according to the heat of each information type on the local platform or across platforms. After the mask is carried out on the space-time matrix, a mask matrix of the space-time matrix can be obtained.
Therefore, the implicit information matrix can be multiplied by the space-time matrix, and the personalized recommendation matrix of the target user can be obtained.
Further, before the personalized recommendation matrix of the target user is obtained based on the implicit information matrix and the mask matrix, dimension change is performed on the mask matrix, so that the dimension of the mask matrix is the same as that of the implicit information matrix. The dimension of the implicit information matrix is m × k, and the dimension of the mask matrix (which is the same as the dimension of the spatio-temporal matrix) is l × l, so the mask matrix needs to be transformed into k × z, and z may be any positive integer. Matrix multiplication is performed by an implicit information matrix and a mask matrix so that (m × k) × (k × z) ═ m × z.
In one possible embodiment, z may be equal to k.
Further, the dimension transformation of the space-time matrix may be based on the size of l and k, and when l > k, k nearest time points and positions may be truncated. When k > l, then l can be filled, the value of the filling is 0, the number of lines of the filling is k-l lines, and the time point and the position of the filling to the farthest are shown in the following table 5:
TABLE 5
| D1 | D2 | … | Dk-1 | Dk | |
| T1 | 0.80 | … | |||
| T2 | 0.75 | … | |||
| … | … | … | … | … | … |
| Tk-1 | … | 0.32 | |||
| Tk | … | 0.00 |
Therefore, the personalized recommendation matrix of the target user can be obtained, and the personalized recommendation matrix comprises recommendation scores corresponding to k time points or positions.
In a possible embodiment, the mask matrix may be dimensionally changed by a predetermined full convolution network, and the output dimension of the full convolution network is the same according to the dimension of the implicit information matrix. Specifically, the mask matrix may be regarded as a graph and used as an input of the full convolution network, the mask matrix is calculated according to parameters of the full convolution network, a dimension of the reconstructed mask matrix is k × z, in this case, z may be equal to l through transformation of the full convolution network, and the reconstructed mask matrix is no longer a diagonal matrix but a time matrix and a space matrix are globally distributed in a k × l matrix, and the implicit information matrix is matrix-multiplied by the mask matrix, so that (m × k) × (k × l) ═ m × l. Thus, the personalized recommendation matrix retains recommendation scores corresponding to the i time points or positions.
The personalized recommendation matrix of the target user may be a time-point-based personalized recommendation matrix or a location-based personalized recommendation matrix, and specifically, depending on the selection of the user, the user may select a corresponding time mode or a location mode on the interactive interface, where the time mode corresponds to the time-point-based personalized recommendation matrix and the location mode corresponds to the location-based personalized recommendation matrix.
Further, the time-based personalized recommendation matrix may be obtained by an implicit information matrix (m × k) × a mask matrix (k × l), and the location-based personalized recommendation matrix may be obtained by an implicit information matrix (m × k) × a mask matrix (l × k)TIt is found that the mask matrix (k × l) indicates that l time points are reserved and the mask matrix (l × k) indicates that l positions are reserved.
Specifically, the time-based personalized recommendation matrix of the target user is shown in table 6 below:
TABLE 6
Specifically, the personalized recommendation matrix based on the location of the target user is shown in the following table 7:
TABLE 7
| T 1 | T 2 | … | T l-1 | Tl | |
| mac1 | x1,1 | x 1,2 | … | x 1,l-1 | x 1,l |
| mac2 | x 2,1 | x 2,2 | … | x 2,l-1 | x 2,l |
| … | … | … | … | … | … |
| macm-1 | x m-1,1 | x m-1,2 | … | x m-1,l-1 | x m-1,l |
| macm | x m,1 | x m,2 | … | x 1,l-1 | x m,l |
In a possible embodiment, the score matrix may be subjected to implicit factorization to obtain an implicit user matrix based on user information, where the implicit user matrix includes the user information and a second implicit score; calculating a mask coefficient of the target user according to the second implicit score; and performing mask calculation on the mask matrix according to the mask coefficient.
The implicit user matrix is shown in table 4, and it should be noted that the implicit matrix includes a second implicit score of the target user, and the second implicit score may be directly used as a mask coefficient and the mask matrix, so as to increase the personalized influence of the target user as an individual user. The second hiding score may also be subjected to weighting calculation to obtain a mask coefficient of the target user, where the weighting of the weighting calculation is a quantization ratio of the browsing information duration of the target user one day (yesterday) to 24 hours, and the longer the browsing information duration of the user one day before, the higher the weighting is, the highest the weighting is 0.9 (for example, the browsing information duration of the user one day before is 24 hours, the optimized ratio is 0.9), and the lowest the weighting is 0.1 (for example, the browsing information duration of the user one day before is 0 hours, the optimized ratio is 0.1).
At this time, the personalized recommendation matrix of the target user is obtained by (m × k) × (k × l) ═ λ (m × l). Wherein λ is a mask coefficient.
105. And carrying out personalized recommendation on the target user according to the personalized recommendation matrix.
In the embodiment of the invention, TOP-K information types with highest information type scores in the personalized recommendation matrix can be selected for recommendation according to different time points or position points, and TOP-N information with highest information score values in the information is selected from the TOP-K information types for recommendation.
In the embodiment of the invention, a time-space matrix of a target user during information browsing is obtained, wherein the time-space matrix comprises browsing time, browsing place and browsing labels, and different browsing labels correspond to different types of information contents; acquiring a scoring matrix of all information, wherein the scoring matrix comprises information tags, user information and scoring values, and different information tags correspond to different types of information contents; carrying out implicit factorization on the scoring matrix to obtain an implicit information matrix based on an information label, wherein the implicit information matrix comprises the information label and a first implicit score; obtaining an individualized recommendation matrix of the target user based on the implicit information matrix and the space-time matrix; and carrying out personalized recommendation on the target user according to the personalized recommendation matrix. The method and the device can recommend the information to the user according to the information of the user in time and space on the basis of information scoring, and fully explores the browsing personality of the user due to the fact that the user is interested in the information when, where and so as to improve the recommendation accuracy.
It should be noted that the method for personalized recommendation based on spatio-temporal features provided in the embodiments of the present invention may be applied to devices such as a mobile phone, a computer, and a server that can perform personalized recommendation based on spatio-temporal features.
In a second aspect, please refer to fig. 3, where fig. 3 is a device for personalized recommendation based on spatiotemporal features according to an embodiment of the present invention, the device includes:
the first obtaining module 301 is configured to obtain a spatio-temporal matrix when a target user browses information, where the spatio-temporal matrix includes browsing time, browsing location, and browsing tags, and different browsing tags correspond to different types of information content;
a second obtaining module 302, configured to obtain a scoring matrix of all information, where the scoring matrix includes information tags, user information, and scoring values, and different information tags correspond to different types of information content;
a decomposition module 303, configured to perform implicit factorization on the score matrix to obtain an implicit information matrix based on an information tag, where the implicit information matrix includes the information tag and a first implicit score;
the processing module 304 is configured to obtain an individualized recommendation matrix of the target user based on the implicit information matrix and the spatio-temporal matrix;
and the recommending module 305 is configured to perform personalized recommendation on the target user according to the personalized recommendation matrix.
Optionally, as shown in fig. 4, the second obtaining module 302 includes:
the obtaining submodule 3021 is configured to obtain historical browsing information of the platform, where the historical browsing information includes an information tag of browsed information, browsed times, positive evaluation times, negative evaluation times, and user information corresponding to the browsed information;
a calculating submodule 3022, configured to calculate a score value of the browsed information according to the browsed times, the positive evaluation times, and the negative evaluation times of the browsed information;
the constructing sub-module 3023 is configured to construct the scoring matrix according to the information tag of the browsed information, the score value of the browsed information, and the user information corresponding to the browsed information.
Optionally, as shown in fig. 5, the processing module 304 includes:
a mask submodule 3041, configured to perform a mask on the space-time matrix to obtain a mask matrix of the space-time matrix, where one mask value corresponds to one browsing tag in the mask matrix;
the processing submodule 3042 is configured to obtain a personalized recommendation matrix of the target user based on the implicit information matrix and the mask matrix.
Optionally, as shown in fig. 6, the processing module 304 further includes:
a converting submodule 3043, configured to perform dimension change on the mask matrix, so that the dimension of the mask matrix is the same as that of the implicit information matrix.
Optionally, the converting submodule 3043 is further configured to perform dimension change on the mask matrix through a preset full convolution network, where output dimensions of the full convolution network are the same according to the dimension of the implicit information matrix.
Optionally, as shown in fig. 7, the mask submodule 3041 includes:
a decomposition unit 30411, configured to perform implicit factorization on the scoring matrix to obtain an implicit user matrix based on user information, where the implicit user matrix includes the user information and a second implicit score;
a calculating unit 30412, configured to calculate a mask coefficient of the target user according to the second implicit score;
a mask unit 30413, configured to perform mask calculation on the mask matrix according to the mask coefficient.
It should be noted that the personalized recommendation device based on spatio-temporal features provided in the embodiments of the present invention may be applied to devices such as mobile phones, computers, servers, etc. that can perform personalized recommendation based on spatio-temporal features.
The personalized recommendation device based on the space-time characteristics provided by the embodiment of the invention can realize each process realized by the personalized recommendation method based on the space-time characteristics in the method embodiment, and can achieve the same beneficial effects. To avoid repetition, further description is omitted here.
Referring to fig. 8, fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present invention, as shown in fig. 8, including: a memory 802, a processor 801, and a computer program stored on the memory 802 and executable on the processor 801, wherein:
the processor 801 is used to call the computer program stored in the memory 802, and executes the following steps:
acquiring a time-space matrix of a target user when browsing information, wherein the time-space matrix comprises browsing time, browsing places and browsing labels, and different browsing labels correspond to different types of information contents;
acquiring a scoring matrix of all information, wherein the scoring matrix comprises information tags, user information and scoring values, and different information tags correspond to different types of information contents;
carrying out implicit factorization on the scoring matrix to obtain an implicit information matrix based on an information label, wherein the implicit information matrix comprises the information label and a first implicit score;
obtaining an individualized recommendation matrix of the target user based on the implicit information matrix and the space-time matrix;
and carrying out personalized recommendation on the target user according to the personalized recommendation matrix.
Optionally, the scoring matrix for acquiring all information executed by the processor 801 includes:
acquiring historical browsing information of a platform, wherein the historical browsing information comprises an information tag of browsed information, browsed times, positive evaluation times, negative evaluation times and user information corresponding to the browsed information;
calculating the score value of the browsed information according to the browsed times, the positive evaluation times and the negative evaluation times of the browsed information;
and constructing the scoring matrix according to the information tag of the browsed information, the scoring value of the browsed information and the user information corresponding to the browsed information.
Optionally, the obtaining, by the processor 801, the personalized recommendation matrix of the target user based on the implicit information matrix and the spatio-temporal matrix includes:
masking the space-time matrix to obtain a mask matrix of the space-time matrix, wherein one mask value corresponds to one browsing label in the mask matrix;
and obtaining the personalized recommendation matrix of the target user based on the implicit information matrix and the mask matrix.
Optionally, before obtaining the personalized recommendation matrix of the target user based on the implicit information matrix and the mask matrix, the obtaining, by the processor 801, the personalized recommendation matrix of the target user based on the implicit information matrix and the spatio-temporal matrix further includes:
and carrying out dimension change on the mask matrix so that the dimension of the mask matrix is the same as that of the implicit information matrix.
Optionally, the performing, by the processor 801, dimension change on the mask matrix includes:
and carrying out dimension change on the mask matrix through a preset full convolution network, wherein the output dimension of the full convolution network is the same according to the dimension of the implicit information matrix.
Optionally, the masking the space-time matrix by the processor 801 to obtain a mask matrix of the space-time matrix includes:
carrying out implicit factorization on the scoring matrix to obtain an implicit user matrix based on user information, wherein the implicit user matrix comprises the user information and a second implicit score;
calculating a mask coefficient of the target user according to the second implicit score;
and performing mask calculation on the mask matrix according to the mask coefficient.
The electronic device may be a device that can be applied to a mobile phone, a computer, a server, and the like that can perform personalized recommendation based on spatiotemporal features.
The electronic device provided by the embodiment of the invention can realize each process realized by the space-time characteristic-based personalized recommendation method in the method embodiment, can achieve the same beneficial effects, and is not repeated here to avoid repetition.
The embodiment of the present invention further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the computer program implements each process of the space-time feature-based personalized recommendation method provided in the embodiment of the present invention, and can achieve the same technical effect, and in order to avoid repetition, the detailed description is omitted here.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
The above disclosure is only for the purpose of illustrating the preferred embodiments of the present invention, and it is therefore to be understood that the invention is not limited by the scope of the appended claims.
Claims (10)
1. A personalized recommendation method based on spatio-temporal characteristics is characterized by comprising the following steps:
acquiring a time-space matrix of a target user when browsing information, wherein the time-space matrix comprises browsing time, browsing places and browsing labels, and different browsing labels correspond to different types of information contents;
acquiring a scoring matrix of all information, wherein the scoring matrix comprises information tags, user information and scoring values, and different information tags correspond to different types of information contents;
carrying out implicit factorization on the scoring matrix to obtain an implicit information matrix based on an information label, wherein the implicit information matrix comprises the information label and a first implicit score;
obtaining an individualized recommendation matrix of the target user based on the implicit information matrix and the space-time matrix;
and carrying out personalized recommendation on the target user according to the personalized recommendation matrix.
2. The method of claim 1, wherein said obtaining a scoring matrix for all information comprises:
acquiring historical browsing information of a platform, wherein the historical browsing information comprises an information tag of browsed information, browsed times, positive evaluation times, negative evaluation times and user information corresponding to the browsed information;
calculating the score value of the browsed information according to the browsed times, the positive evaluation times and the negative evaluation times of the browsed information;
and constructing the scoring matrix according to the information tag of the browsed information, the scoring value of the browsed information and the user information corresponding to the browsed information.
3. The method of claim 2, wherein the deriving the personalized recommendation matrix for the target user based on the implicit information matrix and the spatio-temporal matrix comprises:
masking the space-time matrix to obtain a mask matrix of the space-time matrix, wherein one mask value corresponds to one browsing label in the mask matrix;
and obtaining the personalized recommendation matrix of the target user based on the implicit information matrix and the mask matrix.
4. The method of claim 3, wherein before the obtaining the personalized recommendation matrix for the target user based on the implicit information matrix and the mask matrix, the obtaining the personalized recommendation matrix for the target user based on the implicit information matrix and the spatio-temporal matrix further comprises:
and carrying out dimension change on the mask matrix so that the dimension of the mask matrix is the same as that of the implicit information matrix.
5. The method of claim 4, wherein the dimension changing the mask matrix comprises:
and carrying out dimension change on the mask matrix through a preset full convolution network, wherein the output dimension of the full convolution network is the same according to the dimension of the implicit information matrix.
6. The method of claim 3, wherein said masking said spatio-temporal matrix to obtain a masking matrix of said spatio-temporal matrix comprises:
carrying out implicit factorization on the scoring matrix to obtain an implicit user matrix based on user information, wherein the implicit user matrix comprises the user information and a second implicit score;
calculating a mask coefficient of the target user according to the second implicit score;
and performing mask calculation on the mask matrix according to the mask coefficient.
7. An apparatus for personalized recommendation based on spatiotemporal features, the apparatus comprising:
the system comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring a space-time matrix of a target user when browsing information, the space-time matrix comprises browsing time, browsing place and browsing labels, and different browsing labels correspond to different types of information contents;
the second acquisition module is used for acquiring a scoring matrix of all information, wherein the scoring matrix comprises information tags, user information and scoring values, and different information tags correspond to different types of information contents;
the decomposition module is used for carrying out implicit factorization on the scoring matrix to obtain an implicit information matrix based on an information label, and the implicit information matrix comprises the information label and a first implicit score;
the processing module is used for obtaining the personalized recommendation matrix of the target user based on the implicit information matrix and the space-time matrix;
and the recommending module is used for carrying out personalized recommendation on the target user according to the personalized recommendation matrix.
8. The apparatus of claim 7, wherein the second obtaining module comprises:
the acquisition submodule is used for acquiring historical browsing information of the platform, wherein the historical browsing information comprises an information tag of browsed information, browsed times, positive evaluation times, negative evaluation times and user information corresponding to the browsed information;
the calculation submodule is used for calculating the score value of the browsed information according to the browsed times, the positive evaluation times and the negative evaluation times of the browsed information;
and the construction sub-module is used for constructing the scoring matrix according to the information tag of the browsed information, the scoring value of the browsed information and the user information corresponding to the browsed information.
9. An electronic device, comprising: memory, processor and computer program stored on the memory and executable on the processor, the processor implementing the steps in the spatiotemporal feature-based personalized recommendation method according to any one of claims 1 to 6 when executing the computer program.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the steps in the spatio-temporal feature-based personalized recommendation method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110747365.0A CN113449210B (en) | 2021-07-01 | 2021-07-01 | Personalized recommendation method and device based on space-time characteristics, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110747365.0A CN113449210B (en) | 2021-07-01 | 2021-07-01 | Personalized recommendation method and device based on space-time characteristics, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113449210A true CN113449210A (en) | 2021-09-28 |
| CN113449210B CN113449210B (en) | 2023-01-31 |
Family
ID=77814745
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110747365.0A Active CN113449210B (en) | 2021-07-01 | 2021-07-01 | Personalized recommendation method and device based on space-time characteristics, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113449210B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117556131A (en) * | 2023-10-30 | 2024-02-13 | 广州圈量网络信息科技有限公司 | User tag triggering method, device, equipment and medium based on multiple materials |
Citations (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2011257955A (en) * | 2010-06-08 | 2011-12-22 | Nippon Telegr & Teleph Corp <Ntt> | Recommended item filtering method and recommended item filtering program |
| US20120023051A1 (en) * | 2010-07-22 | 2012-01-26 | Ramin Pishehvar | Signal coding with adaptive neural network |
| US20120191640A1 (en) * | 2011-01-21 | 2012-07-26 | International Business Machines Corporation | Mining temporal patterns in longitudinal event data using discrete event matrices and sparse coding |
| US20150193548A1 (en) * | 2014-01-08 | 2015-07-09 | Rovi Technologies Corporation | Recommendation System With Metric Transformation |
| CN105025091A (en) * | 2015-06-26 | 2015-11-04 | 南京邮电大学 | Shop recommendation method based on position of mobile user |
| CN105808680A (en) * | 2016-03-02 | 2016-07-27 | 西安电子科技大学 | Tensor decomposition based context-dependent position recommendation method |
| CN106296337A (en) * | 2016-07-13 | 2017-01-04 | 扬州大学 | Dynamic recommendation method based on Non-negative Matrix Factorization |
| CN106779867A (en) * | 2016-12-30 | 2017-05-31 | 中国民航信息网络股份有限公司 | Support vector regression based on context-aware recommends method and system |
| CN107391582A (en) * | 2017-06-21 | 2017-11-24 | 浙江工商大学 | The information recommendation method of user preference similarity is calculated based on context ontology tree |
| US20180048530A1 (en) * | 2015-10-23 | 2018-02-15 | Nec Europe Ltd. | Method and system for supporting detection of irregularities in a network |
| US20180075353A1 (en) * | 2016-09-13 | 2018-03-15 | Sap Se | Method and system for cold start video recommendation |
| CN108921670A (en) * | 2018-07-04 | 2018-11-30 | 重庆大学 | A kind of potential interest of fusion user, the Drug trading recommended method of space-time data and classification popularity |
| CN109670909A (en) * | 2018-12-13 | 2019-04-23 | 南京财经大学 | A kind of travelling products recommended method decomposed based on probability matrix with Fusion Features |
| CN110069713A (en) * | 2019-04-24 | 2019-07-30 | 南京邮电大学 | A kind of personalized recommendation method based on user's context perception |
| CN110705996A (en) * | 2019-10-17 | 2020-01-17 | 支付宝(杭州)信息技术有限公司 | User behavior identification method, system and device based on feature mask |
| CN111125517A (en) * | 2019-12-06 | 2020-05-08 | 陕西师范大学 | Implicit matrix decomposition recommendation method based on differential privacy and time perception |
| CN111144765A (en) * | 2019-12-27 | 2020-05-12 | 深圳集智数字科技有限公司 | Method and device for determining incidence relation between specific places |
| CN111737427A (en) * | 2020-05-11 | 2020-10-02 | 华南理工大学 | Mu lesson forum post recommendation method integrating forum interaction behavior and user reading preference |
| CN112818216A (en) * | 2021-01-13 | 2021-05-18 | 平安科技（深圳）有限公司 | Client recommendation method and device, electronic equipment and storage medium |
-
2021
- 2021-07-01 CN CN202110747365.0A patent/CN113449210B/en active Active
Patent Citations (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2011257955A (en) * | 2010-06-08 | 2011-12-22 | Nippon Telegr & Teleph Corp <Ntt> | Recommended item filtering method and recommended item filtering program |
| US20120023051A1 (en) * | 2010-07-22 | 2012-01-26 | Ramin Pishehvar | Signal coding with adaptive neural network |
| US20120191640A1 (en) * | 2011-01-21 | 2012-07-26 | International Business Machines Corporation | Mining temporal patterns in longitudinal event data using discrete event matrices and sparse coding |
| US20150193548A1 (en) * | 2014-01-08 | 2015-07-09 | Rovi Technologies Corporation | Recommendation System With Metric Transformation |
| CN105025091A (en) * | 2015-06-26 | 2015-11-04 | 南京邮电大学 | Shop recommendation method based on position of mobile user |
| US20180048530A1 (en) * | 2015-10-23 | 2018-02-15 | Nec Europe Ltd. | Method and system for supporting detection of irregularities in a network |
| CN105808680A (en) * | 2016-03-02 | 2016-07-27 | 西安电子科技大学 | Tensor decomposition based context-dependent position recommendation method |
| CN106296337A (en) * | 2016-07-13 | 2017-01-04 | 扬州大学 | Dynamic recommendation method based on Non-negative Matrix Factorization |
| US20180075353A1 (en) * | 2016-09-13 | 2018-03-15 | Sap Se | Method and system for cold start video recommendation |
| CN106779867A (en) * | 2016-12-30 | 2017-05-31 | 中国民航信息网络股份有限公司 | Support vector regression based on context-aware recommends method and system |
| CN107391582A (en) * | 2017-06-21 | 2017-11-24 | 浙江工商大学 | The information recommendation method of user preference similarity is calculated based on context ontology tree |
| CN108921670A (en) * | 2018-07-04 | 2018-11-30 | 重庆大学 | A kind of potential interest of fusion user, the Drug trading recommended method of space-time data and classification popularity |
| CN109670909A (en) * | 2018-12-13 | 2019-04-23 | 南京财经大学 | A kind of travelling products recommended method decomposed based on probability matrix with Fusion Features |
| CN110069713A (en) * | 2019-04-24 | 2019-07-30 | 南京邮电大学 | A kind of personalized recommendation method based on user's context perception |
| CN110705996A (en) * | 2019-10-17 | 2020-01-17 | 支付宝(杭州)信息技术有限公司 | User behavior identification method, system and device based on feature mask |
| CN111125517A (en) * | 2019-12-06 | 2020-05-08 | 陕西师范大学 | Implicit matrix decomposition recommendation method based on differential privacy and time perception |
| CN111144765A (en) * | 2019-12-27 | 2020-05-12 | 深圳集智数字科技有限公司 | Method and device for determining incidence relation between specific places |
| CN111737427A (en) * | 2020-05-11 | 2020-10-02 | 华南理工大学 | Mu lesson forum post recommendation method integrating forum interaction behavior and user reading preference |
| CN112818216A (en) * | 2021-01-13 | 2021-05-18 | 平安科技（深圳）有限公司 | Client recommendation method and device, electronic equipment and storage medium |
Non-Patent Citations (7)
| Title |
|---|
| 彭鹏等: "基于用户信任和兴趣的概率矩阵分解推荐方法", 《计算机系统应用》 * |
| 李春春等: "基于ASVD的协同过滤推荐算法", 《小型微型计算机系统》 * |
| 秦晓晖: "基于协同过滤的个性化微博推荐算法研究", 《软件工程》 * |
| 程盼: "面向移动流媒体的缓存预取技术研究", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 * |
| 郑麟等: "基于属性提升与局部采样的推荐评分预测", 《计算机学报》 * |
| 顾军华等: "融合标签信息的卷积矩阵分解推荐算法", 《计算机应用与软件》 * |
| 黄立威等: "基于深度学习的推荐系统研究综述", 《计算机学报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117556131A (en) * | 2023-10-30 | 2024-02-13 | 广州圈量网络信息科技有限公司 | User tag triggering method, device, equipment and medium based on multiple materials |
| CN117556131B (en) * | 2023-10-30 | 2024-04-19 | 广州圈量网络信息科技有限公司 | User tag triggering method, device, equipment and medium based on multiple materials |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113449210B (en) | 2023-01-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20170329856A1 (en) | Method and device for selecting data content to be pushed to terminal, and non-transitory computer storage medium | |
| CN112364204A (en) | Video searching method and device, computer equipment and storage medium | |
| CN107247728B (en) | Text processing method and device and computer storage medium | |
| CN110866042A (en) | Intelligent table query method and device and computer readable storage medium | |
| CN111400551B (en) | Video classification method, electronic equipment and storage medium | |
| CN112070550A (en) | Keyword determination method, device and equipment based on search platform and storage medium | |
| CN113449210B (en) | Personalized recommendation method and device based on space-time characteristics, electronic equipment and storage medium | |
| CN114091472B (en) | Training method of multi-label classification model | |
| CN108154288A (en) | New user subsidizes computational methods, device, electronic equipment and the storage medium of strategy | |
| CN112765450A (en) | Recommended content determining method, recommended content determining device and storage medium | |
| CN116703682A (en) | Government affair data platform based on deep learning | |
| CN115730217A (en) | Model training method, material recalling method and device | |
| CN110263134B (en) | Intelligent emotion question-answering method and device and computer readable storage medium | |
| CN116610853A (en) | Search recommendation method, search recommendation system, computer device, and storage medium | |
| CN114648379A (en) | Recommended commodity determining method and device, electronic equipment and readable storage medium | |
| CN116522917B (en) | Public opinion information popularity scoring method, public opinion information popularity scoring device, computer equipment and storage medium | |
| CN111291263B (en) | Data pushing method, device, system and storage medium | |
| CN117648497B (en) | Method and system for realizing intelligent acquisition of user information based on big data | |
| CN108734549B (en) | Item recommendation method and device | |
| CN116226523B (en) | Knowledge training method and system based on big data | |
| CN113794900B (en) | Video processing method and device | |
| CN115953255A (en) | Service development method, device and equipment based on user obstacle | |
| CN113506140A (en) | Enterprise query application advertisement recommendation method and related products | |
| US20200226188A1 (en) | Adaptive recommendation system and method for network-based content | |
| CN117951364A (en) | Training of information recommendation model, information recommendation method and related equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |