A method and device for pushing item information

By generating correlation diagrams and calculating item recommendation values ​​using metric learning models, real-time push item information is achieved, solving the real-time and calculation overhead of the existing recommendation system, and improving the accuracy and efficiency of recommendations.

CN112116426BActive Publication Date: 2024-11-12CHINA CONSTRUCTION BANK
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010997095.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-09-21
Publication Date
2024-11-12
Estimated Expiration
2040-09-21

AI Technical Summary

Technical Problem

The existing recommendation system cannot push item information in real time, and the calculation overhead is high, resulting in outdated recommendations and waste of resources.

Method used

By obtaining user historical behavior data, generating correlation diagrams and extracting graph features, metric learning models to calculate item recommendation values, real-time recommendations are achieved.

Benefits of technology

It solves the problem of large overhead of real-time push and calculation, improves the accuracy and efficiency of recommendations, reduces the difficulty of model training, and can analyze massive data to identify more features, avoiding outdated recommendations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112116426B_ABST
    Figure CN112116426B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and device for pushing item information, and relates to the field of computer technology. A specific implementation of the method includes: obtaining historical behavior data of each user; generating a correlation graph based on the historical behavior data of each user; extracting graph features in the correlation graph, and calculating the recommendation value of each item for the target user based on the graph features; and pushing item information to the target user based on the recommendation value of each item for the target user. This implementation can solve the technical problems of being unable to push in real time and huge computing overhead.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and device for pushing item information. Background Art

[0002] Recommendation systems are currently widely used in various e-commerce applications. Recommendation technology converts customer data and activities into action intelligence to provide users with item recommendation support. Currently, the more commonly used recommendation technologies include demographic-based recommendations, content-based recommendations, collaborative filtering-based recommendations, etc. Collaborative filtering algorithms include neighborhood-based algorithms (also known as memory-based methods), latent semantic models, and graph-based random walk algorithms.

[0003] In the process of implementing the present invention, the inventors found that there are at least the following problems in the prior art:

[0004] Content-based recommendation solves the cold start problem of items, but it cannot solve the cold start problem of users. It also has the problem of overfitting, and it requires a high level of domain knowledge. It has poor versatility and portability. Changing the form of an item often requires rebuilding a new set. It is also difficult to extract information features from multimedia files, and it can often only be done through manual standard information. Neighborhood-based collaborative filtering algorithms also have problems such as cold start and data sparsity. Model-based collaborative filtering algorithms have solved some of the problems faced by neighborhood-based recommendation algorithms to a certain extent, but the algorithms are usually complex and computationally expensive.

[0005] Due to the limitations of the current system, the above recommendation algorithms all use offline calculations. These recommendations are calculated in advance in the background and cannot use the latest historical data to perform real-time on-demand analysis, resulting in the loss of many recommendation opportunities or outdated recommendations.

[0006] Moreover, the implementation of these algorithms mostly directly utilizes massive amounts of data for model training or preprocessing. That is, they need to train the collected data into models and learn the corresponding rules to generate corresponding models in order to make corresponding predictions for subsequent data. Therefore, the computational overhead of processing is huge. Summary of the invention

[0007] In view of this, an embodiment of the present invention provides a method and device for pushing item information to solve the technical problems of being unable to push in real time and having huge computational overhead.

[0008] To achieve the above object, according to one aspect of an embodiment of the present invention, a method for pushing item information is provided, comprising:

[0009] Get historical behavior data of each user;

[0010] Generate a correlation graph according to the historical behavior data of each user;

[0011] Extracting graph features from the association graph, and calculating the recommendation value of each item to the target user according to the graph features;

[0012] According to the recommendation value of each item to the target user, the item information is pushed to the target user.

[0013] Optionally, the nodes in the association graph represent users and items, and the edges in the association graph represent behavioral relationships between users and items, and the behavioral relationships include at least one of purchase, browsing, and collection.

[0014] Optionally, the historical behavior data includes the user's operation behavior on the item, and the operation behavior includes at least one of purchase, browsing and collection.

[0015] Optionally, generating a correlation graph according to the historical behavior data of each user includes:

[0016] The historical behavior data of each user is subjected to real-time deep correlation analysis technology through a graph database to generate a correlation graph.

[0017] Optionally, the graph database is Neo4J.

[0018] Optionally, calculating the recommendation value of each item to the target user according to the graph features includes:

[0019] The metric learning model is used to learn the graph features, so as to calculate the recommendation value of each item to the target user.

[0020] Optionally, a metric learning model is used to learn the graph features, so as to calculate the recommendation value of each item to the target user, including:

[0021] The graph features are used as input data and input into the metric learning model, and further cluster regression analysis is performed to calculate the recommendation value of each item to the target user.

[0022] Optionally, the graph features are used as input data and input into a metric learning model, and further cluster regression analysis is performed to calculate the recommendation value of each item for the target user, including:

[0023] The metric function formula is used to calculate the recommendation value of each item in the same interest recommendation group for the target user, the recommendation value of each item in the same category recommendation group for the target user, and the recommendation value of each item in the possible interest recommendation group for the target user.

[0024] Optionally, a metric function formula is used to calculate the recommendation value of each item in the same interest recommendation group to the target user, including:

[0025] The metric function formula is used to calculate the recommendation value of each item in the same interest recommendation group to the target user based on the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, and the total number of users.

[0026] Optionally, a metric function formula is used, and according to the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, and the total number of users, the recommendation value of each item in the same interest recommendation group to the target user is calculated, including:

[0027] The following formula is used to calculate the recommendation value of any item in the same interest recommendation group for the target user:

[0028]

[0029]

[0030] Among them, L(f (n,k) ) represents the recommendation value of any item to the target user;

[0031] f (n,k) Represents the coarse-grained recommendation value of any item to the target user;

[0032] K represents the number of steps from any item to the target user, K≠1;

[0033] N represents the total number of users in the association graph;

[0034] f n represents the feature vector of user n, f k The feature vector representing item k.

[0035] Optionally, a metric function formula is used to calculate the recommendation value of each item in the same category recommendation group to the target user or the recommendation value of each item in the possibly-interested recommendation group to the target user, including:

[0036] The metric function formula is used to calculate the recommendation value of each item in the same category recommendation group for the target user or the recommendation value of each item in the possibly interested recommendation group for the target user based on the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, the total number of users, the number of items associated with the target user, the number of users associated with each item, and the total number of items in the same category.

[0037] Optionally, a metric function formula is used, and based on the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, the total number of users, the number of items associated with the target user, the number of users associated with each item, and the total number of items in the same category, the recommendation value of each item in the same category recommendation group to the target user or the recommendation value of each item in the possibly-interested recommendation group to the target user is calculated, including:

[0038] The following formula is used to calculate the recommendation value of any item in the same category recommendation group or in the possibly interested recommendation group for the target user:

[0039]

[0040]

[0041] Among them, L(f (n,k) ) represents the recommendation value of any item to the target user;

[0042] f (n,k) Represents the coarse-grained recommendation value of any item to the target user;

[0043] K represents the number of steps from any item to the target user, K≠1;

[0044] N represents the total number of users in the association graph;

[0045] Ω - Indicates the number of items associated with the target user in the same category recommendation group or in the possibly interested recommendation group in the association graph;

[0046] Ω + Indicates the number of users associated with any item in the same category recommendation group or in the possibly interested recommendation group in the association graph;

[0047] f n represents the feature vector of user n, f k The feature vector representing item k;

[0048] i represents items belonging to the same category, and M refers to the total number of items in the same category.

[0049] Optionally, according to the recommendation value of each item to the target user, pushing item information to the target user includes:

[0050] Sorting the items according to the order of their recommendation values ​​for the target user from large to small, and selecting a number of items at the top of the order;

[0051] Pushing information of the plurality of items to the target user.

[0052] Optionally, the number of the several objects is the same as the number of steps from any one object to the target user.

[0053] Optionally, the number of steps from any one of the items to the target user is preset.

[0054] Optionally, pushing information of the plurality of items to the target user includes:

[0055] According to the historical behavior data of the target user, filtering out operated items from the plurality of items;

[0056] Push the filtered item information to the target user.

[0057] In addition, according to another aspect of an embodiment of the present invention, a device for pushing item information is provided, including:

[0058] The acquisition module is used to obtain the historical behavior data of each user;

[0059] A generating module, used for generating a correlation graph according to the historical behavior data of each user;

[0060] A calculation module, used to extract graph features from the association graph and calculate the recommendation value of each item to the target user according to the graph features;

[0061] The push module is used to push item information to the target user according to the recommendation value of each item to the target user.

[0062] Optionally, the nodes in the association graph represent users and items, and the edges in the association graph represent behavioral relationships between users and items, and the behavioral relationships include at least one of purchase, browsing, and collection.

[0063] Optionally, the historical behavior data includes the user's operation behavior on the item, and the operation behavior includes at least one of purchase, browsing and collection.

[0064] Optionally, the generating module is further used for:

[0065] The historical behavior data of each user is subjected to real-time deep correlation analysis technology through a graph database to generate a correlation graph.

[0066] Optionally, the graph database is Neo4J.

[0067] Optionally, the calculation module is further used for:

[0068] The metric learning model is used to learn the graph features, so as to calculate the recommendation value of each item to the target user.

[0069] Optionally, the calculation module is further used for:

[0070] The graph features are used as input data and input into the metric learning model, and further cluster regression analysis is performed to calculate the recommendation value of each item to the target user.

[0071] Optionally, the calculation module is further used for:

[0072] The metric function formula is used to calculate the recommendation value of each item in the same interest recommendation group for the target user, the recommendation value of each item in the same category recommendation group for the target user, and the recommendation value of each item in the possible interest recommendation group for the target user.

[0073] Optionally, the calculation module is further used for:

[0074] The metric function formula is used to calculate the recommendation value of each item in the same interest recommendation group to the target user based on the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, and the total number of users.

[0075] Optionally, the calculation module is further used for:

[0076] The following formula is used to calculate the recommendation value of any item in the same interest recommendation group for the target user:

[0077]

[0078]

[0079] Among them, L(f (n,k) ) represents the recommendation value of any item to the target user;

[0080] f (n,k) Represents the coarse-grained recommendation value of any item to the target user;

[0081] K represents the number of steps from any item to the target user, K≠1;

[0082] N represents the total number of users in the association graph;

[0083] f n represents the feature vector of user n, f k The feature vector representing item k.

[0084] Optionally, the calculation module is further used for:

[0085] The metric function formula is used to calculate the recommendation value of each item in the same category recommendation group for the target user or the recommendation value of each item in the possibly interested recommendation group for the target user based on the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, the total number of users, the number of items associated with the target user, the number of users associated with each item, and the total number of items in the same category.

[0086] Optionally, the calculation module is further used for:

[0087] The following formula is used to calculate the recommendation value of any item in the same category recommendation group or in the possibly interested recommendation group for the target user:

[0088]

[0089]

[0090] Among them, L(f (n,k) ) represents the recommendation value of any item to the target user;

[0091] f (n,k) Represents the coarse-grained recommendation value of any item to the target user;

[0092] K represents the number of steps from any item to the target user, K≠1;

[0093] N represents the total number of users in the association graph;

[0094] Ω - Indicates the number of items associated with the target user in the same category recommendation group or in the possibly interested recommendation group in the association graph;

[0095] Ω + Indicates the number of users associated with any item in the same category recommendation group or in the possibly interested recommendation group in the association graph;

[0096] f n represents the feature vector of user n, f k The feature vector representing item k;

[0097] i represents items belonging to the same category, and M refers to the total number of items in the same category.

[0098] Optionally, the push module is further used for:

[0099] Sorting the items according to the order of their recommendation values ​​for the target user from large to small, and selecting a number of items at the top of the order;

[0100] Pushing information of the plurality of items to the target user.

[0101] Optionally, the number of the several objects is the same as the number of steps from any one object to the target user.

[0102] Optionally, the number of steps from any one of the items to the target user is preset.

[0103] Optionally, the push module is further used for:

[0104] According to the historical behavior data of the target user, filtering out operated items from the plurality of items;

[0105] Push the filtered item information to the target user.

[0106] According to another aspect of an embodiment of the present invention, there is further provided an electronic device, including:

[0107] one or more processors;

[0108] a storage device for storing one or more programs,

[0109] When the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any one of the above embodiments.

[0110] According to another aspect of an embodiment of the present invention, a computer-readable medium is provided, on which a computer program is stored. When the program is executed by a processor, the method described in any one of the above embodiments is implemented.

[0111] One embodiment of the above invention has the following advantages or beneficial effects: because it adopts a technical means of generating an association graph according to the historical behavior data of each user, extracting the graph features in the association graph, and calculating the recommendation value of each item to the target user according to the graph features, it overcomes the technical problems of the inability to push in real time and the huge computational overhead in the prior art. The embodiment of the present invention utilizes graph database analysis technology to optimize the metric learning algorithm to achieve real-time item recommendation, and solves the problems of the overall algorithm module being heavy, the amount of calculation being large, and the overhead being large in the prior recommendation technology. The embodiment of the present invention can directly perform deep path analysis on the user's behavior data and extract the corresponding graph features, and then combined with the metric learning algorithm, it can greatly reduce the difficulty of model training and improve the accuracy of model prediction; it can identify and consider more features, and can further analyze massive data, and the graph database technology can easily achieve real-time analysis, solving the problem of outdated recommendations in the existing recommendation algorithm.

[0112] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0113] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention.

[0114] Figure 1 is a schematic diagram of the main process of the method for pushing item information according to an embodiment of the present invention;

[0115] Figure 2 is a schematic diagram of an association graph according to an embodiment of the present invention;

[0116] Figure 3 is a schematic diagram of the main process of a method for pushing item information according to a reference embodiment of the present invention;

[0117] Figure 4 is a schematic diagram of the main process of a method for pushing item information according to another reference embodiment of the present invention;

[0118] Figure 5 is a schematic diagram of main modules of an apparatus for pushing item information according to an embodiment of the present invention;

[0119] Figure 6 is an exemplary system architecture diagram to which embodiments of the present invention may be applied;

[0120] Figure 7 It is a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0121] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0122] Figure 1 Schematic diagram of the main process of the method for pushing item information according to an embodiment of the present invention. Figure 1 As shown, the method for pushing item information may include:

[0123] Step 101, obtaining historical behavior data of each user.

[0124] The historical behavior data of each user can be collected through the Restful interface. This interface has the ability to connect with various e-commerce platforms to obtain the user's historical behavior data and send the user's historical behavior data to the graph database in real time. As the input of the original information, the graph database will further perform in-depth correlation analysis.

[0125] In the embodiments of the present invention, the historical behavior data may be behavior data within one month from the current time, behavior data within one week from the current time, or real-time behavior data. These user behaviors that have occurred can be used as historical behavior data, so that the latest historical behavior data can be used to push to the user in real time. Optionally, the historical behavior data includes the user's operation behavior on the item, and the operation behavior includes at least one of purchase, browsing and collection.

[0126] Step 102: Generate a correlation graph based on the historical behavior data of each user.

[0127] In this step, a correlation graph is generated based on the historical behavior data of each user obtained in step 101; wherein the nodes in the correlation graph represent users and items, and the edges in the correlation graph represent the behavior relationship between users and items, and the behavior relationship includes at least one of purchase, browsing and collection. Figure 2 As shown in the figure, the nodes in the association graph represent users and items, such as user nodes and item nodes; the edges between user nodes and item nodes represent the user's behavioral relationship with items, such as a user purchased an item, a user browsed an item, a user collected an item, etc.

[0128] Optionally, step 102 may include: performing real-time deep correlation analysis technology on the historical behavior data of each user through a graph database, thereby generating a correlation graph. In an embodiment of the present invention, a real-time deep correlation analysis is performed on the historical behavior data of the user collected through a Restful interface through a graph database to obtain the behavioral relationship between the user and the item. Optionally, the graph database is Neo4J.

[0129] Step 103: extract graph features from the association graph, and calculate the recommendation value of each item to the target user based on the graph features.

[0130] Optionally, calculating the recommendation value of each item to the target user according to the graph feature includes: using a metric learning model to learn the graph feature, thereby calculating the recommendation value of each item to the target user. Graph feature learning mainly obtains the relationship between users and items, items and items, and users and users, and then uses a metric learning model (MetricLearning) to learn the graph features, thereby calculating the recommendation value of each item to the target user.

[0131] Optionally, a metric learning model is used to learn the graph features, so as to calculate the recommendation value of each item for the target user, including: inputting the graph features as input data into the metric learning model, and further performing cluster regression analysis, so as to calculate the recommendation value of each item for the target user. In an embodiment of the present invention, the graph features are input into the metric learning model, and the metric learning model further performs cluster regression analysis on the graph features, so as to calculate the recommendation value of each item for the target user.

[0132] Optionally, the graph features are used as input data and input into a metric learning model, and further cluster regression analysis is performed to calculate the recommendation value of each item for the target user, including: using a metric function formula to calculate the recommendation value of each item in the same interest recommendation group for the target user, the recommendation value of each item in the same category recommendation group for the target user, and the recommendation value of each item in the possible interest recommendation group for the target user. Figure 2 As shown, the metric function formula can be used to calculate the recommendation value of each item in the same interest recommendation group to the target user, the recommendation value of each item in the same category recommendation group to the target user, and the recommendation value of each item in the possible interest recommendation group to the target user. In this way, item information can be pushed to the target user based on the same interest recommendation group, the same category recommendation group, and the possible interest recommendation group.

[0133] Optionally, a metric function formula is used to calculate the recommendation value of each item in the same interest recommendation group for the target user, including: using the metric function formula, and calculating the recommendation value of each item in the same interest recommendation group for the target user based on the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, and the total number of users. In an embodiment of the present invention, it is also necessary to obtain the feature vector of each item (i.e., the high-order feature value of each item) and the feature vector of the target user (i.e., the high-order feature value of the target user), and then use the metric function formula, and calculate the recommendation value of each item in the same interest recommendation group for the target user based on the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, and the total number of users.

[0134] Optionally, a metric function formula is used, and according to the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, and the total number of users, the recommendation value of each item in the same interest recommendation group to the target user is calculated, including: using the following formula to calculate the recommendation value of any item in the same interest recommendation group to the target user:

[0135]

[0136]

[0137] Among them, L(f (n,k) ) represents the recommendation value of any item to the target user;

[0138] f (n,k) Represents the coarse-grained recommendation value of any item to the target user;

[0139] K represents the number of steps from any item to the target user, K≠1;

[0140] N represents the total number of users in the association graph;

[0141] f n represents the feature vector of user n, f k The feature vector representing item k.

[0142] It should be pointed out that the number of steps from any item to the target user can be preset. If K=3, then the item is any item within three steps from the target user. The fewer the number of steps from the user to the item, the greater its recommendation value should be. The value of K here is mainly determined by the top K items to be recommended in the required ranking, and it is best to keep it consistent with them. Represents the cumulative purchase value of item k by all users within the feature graph.

[0143] Optionally, the recommendation value of each item in the same category recommendation group to the target user or the recommendation value of each item in the possibly interested recommendation group to the target user is calculated by using the metric function formula, and according to the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, the total number of users, the number of items associated with the target user, the number of users associated with each item, and the total number of items in the same category, the recommendation value of each item in the same category recommendation group to the target user or the recommendation value of each item in the possibly interested recommendation group to the target user is calculated. In an embodiment of the present invention, it is also necessary to obtain the feature vector of each item (i.e., the high-order feature value of each item) and the feature vector of the target user (i.e., the high-order feature value of the target user), and then the metric function formula is used, and according to the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, the total number of users, the number of items associated with the target user, the number of users associated with each item, and the total number of items in the same category, the recommendation value of each item in the same category recommendation group to the target user or the recommendation value of each item in the possibly interested recommendation group to the target user is calculated.

[0144] Optionally, a metric function formula is used, and according to the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, the total number of users, the number of items associated with the target user, the number of users associated with each item, and the total number of items in the same category, a recommendation value of each item in the same category recommendation group for the target user or a recommendation value of each item in the possibly-interested recommendation group for the target user is calculated, including: using the following formula to calculate the recommendation value of any item in the same category recommendation group or the possibly-interested recommendation group for the target user:

[0145]

[0146]

[0147] Among them, L(f (n,k) ) represents the recommendation value of any item to the target user;

[0148] f (n,k) Represents the coarse-grained recommendation value of any item to the target user;

[0149] K represents the number of steps from any item to the target user, K≠1;

[0150] N represents the total number of users in the association graph;

[0151] Ω - Indicates the number of items associated with the target user in the same category recommendation group or in the possibly interested recommendation group in the association graph;

[0152] Ω + Indicates the number of users associated with any item in the same category recommendation group or in the possibly interested recommendation group in the association graph;

[0153] f n represents the feature vector of user n, f k The feature vector representing item k;

[0154] i represents items belonging to the same category, and M refers to the total number of items in the same category.

[0155] It should be pointed out that the number of steps from any item to the target user can be preset. If K=5, then the item is any item within five steps from the target user. The fewer the number of steps from the user to the item, the greater its recommendation value should be. The value of K here is mainly determined by the top K items to be recommended in the required ranking, and it is best to keep it consistent with them. Represents the cumulative purchase value of item k by all users within the feature graph.

[0156] Indicates that k takes values ​​from 1 to |Ω- |, all Accumulate to obtain the cumulative purchase value of user n for item categories within the feature graph range.

[0157] Indicates that n ranges from 1 to N. The accumulated purchase value of a certain item category by all users is then obtained. Indicates the number of times item i is purchased.

[0158] Step 104: Push item information to the target user based on the recommendation value of each item to the target user.

[0159] Optionally, step 104 may include: sorting the items in descending order of their recommendation values ​​for the target user, selecting a number of items at the top of the sorting, and pushing information about the items to the target user. Optionally, the number of items is the same as the number of steps from any one item to the target user. After obtaining the recommendation values ​​of the items for the target user, the items may be sorted in descending order of their recommendation values, thereby selecting the top K items in the sorting, and then pushing information about the K items to the target user.

[0160] Optionally, pushing the information of the plurality of items to the target user includes: filtering out the operated items from the plurality of items according to the historical behavior data of the target user; and pushing the information of the filtered items to the target user. If an item that the target user has purchased (i.e., an operated item) appears in the K items, the item can be filtered out from the K items, and only the information of the items that the target user has not purchased can be pushed to the target user, such as Figure 2 shown.

[0161] According to the various embodiments described above, it can be seen that the embodiments of the present invention solve the technical problems of the inability to push in real time and the huge computational overhead in the prior art by generating an association graph based on the historical behavior data of each user, extracting graph features in the association graph, and calculating the recommendation value of each item to the target user based on the graph features. The embodiments of the present invention utilize graph database analysis technology to optimize the metric learning algorithm to achieve real-time item recommendation and solve the problems of the overall algorithm module being heavy, the amount of computation being large, and the overhead being high in the prior art recommendation technology. The embodiments of the present invention can directly perform deep path analysis on the user's behavior data and extract the corresponding graph features. Combined with the metric learning algorithm, the difficulty of model training can be greatly reduced and the accuracy of model prediction can be improved. It can identify and consider more features, further analyze massive data, and the graph database technology can easily achieve real-time analysis, solving the problem of outdated recommendations in the prior art recommendation algorithms.

[0162] Figure 3 is a schematic diagram of the main process of a method for pushing item information according to a reference embodiment of the present invention. As another embodiment of the present invention, Figure 3 As shown, the method for pushing item information may include:

[0163] Step 301, obtaining the historical behavior data of each user, the feature vector of each user, and the feature vector of each item.

[0164] Step 302: Perform real-time deep correlation analysis on the historical behavior data of each user through a graph database to generate a correlation graph.

[0165] Step 303, using a metric function formula, and according to the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, and the total number of users, the recommendation value of each item in the same interest recommendation group to the target user is calculated.

[0166] Step 304: sort the items in descending order of their recommendation values ​​for the target user, and select a number of items at the top of the sort.

[0167] Step 305: Push information of the plurality of items to the target user.

[0168] In addition, the specific implementation content of the method for pushing item information in a reference embodiment of the present invention has been described in detail in the method for pushing item information described above, so the repeated content will not be described here.

[0169] Figure 4 FIG. 1 is a schematic diagram of the main process of a method for pushing item information according to another reference embodiment of the present invention. As another embodiment of the present invention, Figure 4 As shown, the method for pushing item information may include:

[0170] Step 401, obtaining the historical behavior data of each user, the feature vector of each user, and the feature vector of each item.

[0171] Step 402: Perform real-time deep correlation analysis on the historical behavior data of each user through a graph database to generate a correlation graph.

[0172] Step 403, using a metric function formula, and based on the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, the total number of users, the number of items associated with the target user, the number of users associated with each item, and the total number of items in the same category, calculate the recommendation value of each item in the same category recommendation group for the target user or the recommendation value of each item in the possibly interested recommendation group for the target user.

[0173] Step 404: sort the items in descending order of their recommendation values ​​for the target user, and select a number of items at the top of the sort.

[0174] Step 405: Filter out operated items from the plurality of items according to the historical behavior data of the target user.

[0175] Step 406: Push the filtered item information to the target user.

[0176] In addition, the specific implementation content of the method for pushing item information in another reference embodiment of the present invention has been described in detail in the method for pushing item information described above, so the repeated content will not be described here.

[0177] Figure 5 is a schematic diagram of main modules of an apparatus for pushing item information according to an embodiment of the present invention. Figure 5 As shown, the device 500 for pushing item information includes an acquisition module 501, a generation module 502, a calculation module 503 and a push module 504; wherein the acquisition module 501 is used to acquire the historical behavior data of each user; the generation module 502 is used to generate an association graph according to the historical behavior data of each user; the calculation module 503 is used to extract graph features in the association graph, and calculate the recommendation value of each item to the target user according to the graph features; the push module 504 is used to push item information to the target user according to the recommendation value of each item to the target user.

[0178] Optionally, the nodes in the association graph represent users and items, and the edges in the association graph represent behavioral relationships between users and items, and the behavioral relationships include at least one of purchase, browsing, and collection.

[0179] Optionally, the historical behavior data includes the user's operation behavior on the item, and the operation behavior includes at least one of purchase, browsing and collection.

[0180] Optionally, the generating module 502 is further configured to:

[0181] The historical behavior data of each user is subjected to real-time deep correlation analysis technology through a graph database to generate a correlation graph.

[0182] Optionally, the graph database is Neo4J.

[0183] Optionally, the calculation module 503 is further used for:

[0184] The metric learning model is used to learn the graph features, so as to calculate the recommendation value of each item to the target user.

[0185] Optionally, the calculation module 503 is further used for:

[0186] The graph features are used as input data and input into the metric learning model, and further cluster regression analysis is performed to calculate the recommendation value of each item to the target user.

[0187] Optionally, the calculation module 503 is further used for:

[0188] The metric function formula is used to calculate the recommendation value of each item in the same interest recommendation group for the target user, the recommendation value of each item in the same category recommendation group for the target user, and the recommendation value of each item in the possible interest recommendation group for the target user.

[0189] Optionally, the calculation module 503 is further used for:

[0190] The metric function formula is used to calculate the recommendation value of each item in the same interest recommendation group to the target user based on the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, and the total number of users.

[0191] Optionally, the calculation module 503 is further used for:

[0192] The following formula is used to calculate the recommendation value of any item in the same interest recommendation group for the target user:

[0193]

[0194]

[0195] Among them, L(f (n,k) ) represents the recommendation value of any item to the target user;

[0196] f (n,k) Represents the coarse-grained recommendation value of any item to the target user;

[0197] K represents the number of steps from any item to the target user, K≠1;

[0198] N represents the total number of users in the association graph;

[0199] f nrepresents the feature vector of user n, f k The feature vector representing item k.

[0200] Optionally, the calculation module 503 is further used for:

[0201] The metric function formula is used to calculate the recommendation value of each item in the same category recommendation group for the target user or the recommendation value of each item in the possibly interested recommendation group for the target user based on the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, the total number of users, the number of items associated with the target user, the number of users associated with each item, and the total number of items in the same category.

[0202] Optionally, the calculation module 503 is further used for:

[0203] The following formula is used to calculate the recommendation value of any item in the same category recommendation group or in the possibly interested recommendation group for the target user:

[0204]

[0205]

[0206] Among them, L(f (n,k) ) represents the recommendation value of any item to the target user;

[0207] f (n,k) Represents the coarse-grained recommendation value of any item to the target user;

[0208] K represents the number of steps from any item to the target user, K≠1;

[0209] N represents the total number of users in the association graph;

[0210] Ω - Indicates the number of items associated with the target user in the same category recommendation group or in the possibly interested recommendation group in the association graph;

[0211] Ω + Indicates the number of users associated with any item in the same category recommendation group or in the possibly interested recommendation group in the association graph;

[0212] f n represents the feature vector of user n, f k The feature vector representing item k;

[0213] i represents items belonging to the same category, and M refers to the total number of items in the same category.

[0214] Optionally, the push module 504 is further used for:

[0215] Sorting the items according to the order of their recommendation values ​​for the target user from large to small, and selecting a number of items at the top of the order;

[0216] Pushing information of the plurality of items to the target user.

[0217] Optionally, the number of the several objects is the same as the number of steps from any one object to the target user.

[0218] Optionally, the number of steps from any one of the items to the target user is preset.

[0219] Optionally, the push module 504 is further used for:

[0220] According to the historical behavior data of the target user, filtering out operated items from the plurality of items;

[0221] Push the filtered item information to the target user.

[0222] According to the various embodiments described above, it can be seen that the embodiments of the present invention solve the technical problems of the inability to push in real time and the huge computational overhead in the prior art by generating an association graph based on the historical behavior data of each user, extracting graph features in the association graph, and calculating the recommendation value of each item to the target user based on the graph features. The embodiments of the present invention utilize graph database analysis technology to optimize the metric learning algorithm to achieve real-time item recommendation and solve the problems of the overall algorithm module being heavy, the amount of computation being large, and the overhead being high in the prior art recommendation technology. The embodiments of the present invention can directly perform deep path analysis on the user's behavior data and extract the corresponding graph features. Combined with the metric learning algorithm, the difficulty of model training can be greatly reduced and the accuracy of model prediction can be improved. It can identify and consider more features, further analyze massive data, and the graph database technology can easily achieve real-time analysis, solving the problem of outdated recommendations in the prior art recommendation algorithms.

[0223] It should be noted that the specific implementation content of the device for pushing item information of the present invention has been described in detail in the method for pushing item information described above, so the repeated content will not be described again here.

[0224] Figure 6 An exemplary system architecture 600 is shown to which the method for pushing item information or the apparatus for pushing item information according to the embodiments of the present invention may be applied.

[0225] like Figure 6As shown, system architecture 600 may include terminal devices 601, 602, 603, network 604 and server 605. Network 604 is used to provide a medium for communication links between terminal devices 601, 602, 603 and server 605. Network 604 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0226] Users can use terminal devices 601, 602, and 603 to interact with server 605 through network 604 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 601, 602, and 603, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only examples).

[0227] The terminal devices 601 , 602 , and 603 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers, etc.

[0228] The server 605 may be a server that provides various services, such as a backend management server (only an example) that provides support for shopping websites browsed by users using the terminal devices 601, 602, and 603. The backend management server may analyze and process the received data such as the item information query request, and feed back the processing results (such as target push information, item information - only an example) to the terminal device.

[0229] It should be noted that the method for pushing item information provided in the embodiment of the present invention is generally executed by the server 605, and accordingly, the device for pushing item information is generally set in the server 605. The method for pushing item information provided in the embodiment of the present invention can also be executed by the terminal devices 601, 602, and 603, and accordingly, the device for pushing item information can be set in the terminal devices 601, 602, and 603.

[0230] It should be understood that Figure 6 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0231] Reference below Figure 7 , which shows a schematic diagram of the structure of a computer system 700 of a terminal device suitable for implementing an embodiment of the present invention. Figure 7 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0232] like Figure 7As shown, the computer system 700 includes a central processing unit (CPU) 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage part 708 into a random access memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the system 700 are also stored. The CPU 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0233] The following components are connected to the I / O interface 705: an input section 706 including a keyboard, a mouse, etc.; an output section 707 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 708 including a hard disk, etc.; and a communication section 709 including a network interface card such as a LAN card, a modem, etc. The communication section 709 performs communication processing via a network such as the Internet. A drive 710 is also connected to the I / O interface 705 as needed. A removable medium 711, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 710 as needed, so that a computer program read therefrom is installed into the storage section 708 as needed.

[0234] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 709, and / or installed from the removable medium 711. When the computer program is executed by the central processing unit (CPU) 701, the above-mentioned functions defined in the system of the present invention are executed.

[0235] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0236] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, functions and operations of the system, method and computer program according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0237] The modules involved in the embodiments of the present invention may be implemented by software or hardware. The modules described may also be set in a processor, for example, it may be described as: a processor includes an acquisition module, a generation module, a calculation module and a push module, wherein the names of these modules do not constitute a limitation on the modules themselves in some cases.

[0238] As another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiment; or may exist independently without being assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device implements the following method: obtaining historical behavior data of each user; generating an association graph based on the historical behavior data of each user; extracting graph features in the association graph, and calculating the recommendation value of each item for the target user based on the graph features; and pushing item information to the target user based on the recommendation value of each item for the target user.

[0239] According to the technical solution of the embodiment of the present invention, because it adopts a technical means of generating an association graph according to the historical behavior data of each user, extracting the graph features in the association graph, and calculating the recommendation value of each item to the target user according to the graph features, it overcomes the technical problems of the inability to push in real time and the huge computational overhead in the prior art. The embodiment of the present invention uses graph database analysis technology to optimize the metric learning algorithm to achieve real-time item recommendation and solve the problems of the overall algorithm module being heavy, the amount of calculation being large, and the overhead being large in the prior recommendation technology. The embodiment of the present invention can directly perform deep path analysis on the user's behavior data and extract the corresponding graph features. Combined with the metric learning algorithm, it can greatly reduce the difficulty of model training and improve the accuracy of model prediction; it can identify and consider more features, further analyze massive data, and the graph database technology can easily achieve real-time analysis, solving the problem of outdated recommendations in the prior recommendation algorithm.

[0240] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for pushing item information, It is characterized in that include: Get historical behavior data of each user; Generate a correlation graph according to the historical behavior data of each user; Extracting graph features from the association graph, and calculating the recommendation value of each item to the target user according to the graph features; Pushing item information to the target user according to the recommendation value of each item to the target user; The recommendation value of each item to the target user is calculated according to the graph features, including: A metric learning model is used to learn the graph features, thereby calculating the recommendation value of each item to the target user; The metric learning model is used to learn the graph features, so as to calculate the recommendation value of each item to the target user, including: The graph features are used as input data and input into a metric learning model, and cluster regression analysis is further performed to calculate the recommendation value of each item for the target user; The graph features are used as input data and input into the metric learning model, and further cluster regression analysis is performed to calculate the recommendation value of each item for the target user, including: The metric function formula is used to calculate the recommendation value of each item in the same interest recommendation group, the recommendation value of each item in the same category recommendation group, and the recommendation value of each item in the possible interest recommendation group. The metric function formula is used to calculate the recommendation value of each item in the same interest recommendation group to the target user, including: The metric function formula is used to calculate the recommendation value of each item in the same interest recommendation group to the target user based on the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, and the total number of users; The metric function formula is used to calculate the recommendation value of each item in the same interest recommendation group to the target user based on the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, and the total number of users, including: The following formula is used to calculate the recommendation value of any item in the same interest recommendation group for the target user: in, Indicates the recommendation value of any item to the target user; Represents the coarse-grained recommendation value of any item to the target user; K represents the number of steps from any item to the target user. ; Represents the total number of users in the association graph; represents the feature vector of user n, The feature vector representing item k.

2. The method according to claim 1, It is characterized in that The nodes in the association graph represent users and items, and the edges in the association graph represent the behavioral relationship between users and items, and the behavioral relationship includes at least one of purchase, browsing, and collection.

3. The method according to claim 1, It is characterized in that The historical behavior data includes the user's operation behavior on the item, and the operation behavior includes at least one of purchase, browsing and collection.

4. The method according to claim 1, It is characterized in that Generate a correlation graph based on the historical behavior data of each user, including: The historical behavior data of each user is subjected to real-time deep correlation analysis technology through a graph database to generate a correlation graph.

5. The method according to claim 4, It is characterized in that The graph database is Neo4J.

6. The method according to claim 1, It is characterized in that The metric function formula is used to calculate the recommendation value of each item in the same category recommendation group for the target user or the recommendation value of each item in the possibly interested recommendation group for the target user, including: The metric function formula is used to calculate the recommendation value of each item in the same category recommendation group for the target user or the recommendation value of each item in the possibly interested recommendation group for the target user based on the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, the total number of users, the number of items associated with the target user, the number of users associated with each item, and the total number of items in the same category.

7. The method according to claim 6, It is characterized in that The metric function formula is used, and based on the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, the total number of users, the number of items associated with the target user, the number of users associated with each item, and the total number of items in the same category, the recommendation value of each item in the same category recommendation group to the target user or the recommendation value of each item in the possibly interested recommendation group to the target user is calculated, including: The following formula is used to calculate the recommendation value of any item in the same category recommendation group or in the possibly interested recommendation group for the target user: in, Indicates the recommendation value of any item to the target user; Represents the coarse-grained recommendation value of any item to the target user; K represents the number of steps from any item to the target user. ; Represents the total number of users in the association graph; Indicates the number of items associated with the target user in the same category recommendation group or in the possibly interested recommendation group in the association graph; Indicates the number of users associated with any item in the same category recommendation group or in the possibly interested recommendation group in the association graph; represents the feature vector of user n, The feature vector representing item k; i represents items belonging to the same category, and M refers to the total number of items in the same category. 8.. The method according to claim 1, It is characterized in that Pushing item information to the target user according to the recommendation value of each item to the target user includes: Sorting the items according to the order of their recommendation values ​​for the target user from large to small, and selecting a number of items at the top of the order; Pushing information of the plurality of items to the target user.

9. The method according to claim 8, It is characterized in that The number of the plurality of objects is the same as the number of steps from the any one object to the target user.

10. The method according to claim 9, It is characterized in that The number of steps from any one of the items to the target user is preset.

11. The method according to claim 8, It is characterized in that Pushing information of the plurality of items to the target user includes: According to the historical behavior data of the target user, filtering out operated items from the plurality of items; Push the filtered item information to the target user.

12. A device for pushing item information, It is characterized in that include: The acquisition module is used to obtain the historical behavior data of each user; A generating module, used for generating a correlation graph according to the historical behavior data of each user; A calculation module, used to extract graph features from the association graph and calculate the recommendation value of each item to the target user according to the graph features; A push module, used for pushing item information to the target user according to the recommendation value of each item to the target user; The calculation module is also used for: A metric learning model is used to learn the graph features, thereby calculating the recommendation value of each item to the target user; The calculation module is also used for: The graph features are used as input data and input into a metric learning model, and cluster regression analysis is further performed to calculate the recommendation value of each item for the target user; The calculation module is also used for: The metric function formula is used to calculate the recommendation value of each item in the same interest recommendation group, the recommendation value of each item in the same category recommendation group, and the recommendation value of each item in the possible interest recommendation group. The calculation module is also used for: The metric function formula is used to calculate the recommendation value of each item in the same interest recommendation group to the target user based on the number of steps from each item to the target user, the feature vector of each item, the feature vector of the target user, and the total number of users; The calculation module is also used for: The following formula is used to calculate the recommendation value of any item in the same interest recommendation group for the target user: in, Indicates the recommendation value of any item to the target user; Represents the coarse-grained recommendation value of any item to the target user; K represents the number of steps from any item to the target user. ; Represents the total number of users in the association graph; represents the feature vector of user n, The feature vector representing item k.

13. The device according to claim 12, It is characterized in that The generation module is also used for: The historical behavior data of each user is subjected to real-time deep correlation analysis technology through a graph database to generate a correlation graph.

14. An electronic device, It is characterized in that include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 11.

15. A computer readable medium having a computer program stored thereon, It is characterized in that When the program is executed by a processor, the method according to any one of claims 1 to 11 is implemented.