Method, device and storage medium for pushing display information

Abstract

The application discloses a method, device and equipment for pushing display information and a storage medium, and belongs to the technical field of computers. The method comprises the following steps: obtaining an attribute vector of a target account and an attribute vector of an associated object; determining a feature vector of the target account and a feature vector of the associated object based on the attribute vector of the target account, the attribute vector of the associated object, an association relationship between the associated object and the target account, and a feature extraction model; when a display triggering event of the target account is detected, determining a target object displayed to the target account based on the feature vector of the target account and feature vectors of multiple objects, and pushing display information of the target object to the target account. According to the application, the feature vector of the target account determined fuses relevant information of objects liked and objects disliked by the target account, so that the target object determined by using the feature vector is relatively accurate, the conversion rate of the display information pushed is improved, and the use experience of a user is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computers, and in particular, relates to a method and device for pushing display information, and a storage medium. BACKGROUND

[0002] In some applications, we can often receive display information of display objects pushed by an application, for example, in a video application, a video is usually pushed to a user, in a search application, text and image information is usually pushed to a user, in a shopping application, a commodity is usually pushed to a user, and in a friend-making application, another user is pushed to a user. In addition, an object in positive relationship or negative relationship with the user can also be obtained in the application. The positive relationship refers to that the user is interested in the object, and the negative relationship refers to that the user obviously dislikes the object. Taking the object as a commodity for example, the positive relationship is that the user has browsed display information of the commodity or purchased the commodity, and the negative relationship is that the user has deleted display information of the commodity.

[0003] Generally, a method for determining a display object of a user is a collaborative filtering method, which is based on a positive relationship between the user and the object to push, for example, taking pushing a commodity to a user for example: user A and user B are in a positive relationship with commodity C, it can be considered that the interests of user A and user B are similar, when recommending the commodity to user A, commodities in a positive relationship with user B (i.e., commodities purchased or browsed by user B) are determined as a plurality of display objects corresponding to user A, and display information of the display objects is pushed to user A.

[0004] The above method for determining a display object is rough and one-sided, only considering objects in a positive relationship with the user, and not considering objects in a negative relationship with the user. There can be objects of common interest between two users, and there can also be other objects that one user is interested in and the other user dislikes. Therefore, using the above collaborative filtering method for pushing can send display information of objects disliked by the user to the user, which reduces the conversion rate of the display information and reduces the user experience. SUMMARY

[0005] Embodiments of the present application provide a method for pushing display information, which can solve the problem of low conversion rate of display information and low user experience caused by the rough method for determining a display object in the prior art.

[0006] In a first aspect, a method for pushing display information is provided, and the method comprises:

[0007] obtaining an attribute vector of the target account and an attribute vector of an associated object corresponding to the target account, wherein the associated object is an object having an association relationship with the target account in a plurality of objects for display, and the association relationship is a positive association relationship or a negative association relationship;

[0008] determining a feature vector of the target account and a feature vector of the associated object based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and a trained feature extraction model;

[0009] when a display triggering event of the target account is detected, determining a target object for display of the target account based on the currently stored feature vector of the target account and the feature vectors of the plurality of objects, and pushing display information of the target object to the target account.

[0010] In a possible implementation, before the attribute vector of the target account and the attribute vector of the associated object corresponding to the target account are obtained, the method further includes:

[0011] In the plurality of objects for display, obtaining objects having an association relationship with the target account within a preset time period to obtain initial associated objects corresponding to the target account;

[0012] determining a first number of initial associated objects having a positive association relationship with the target account and a second number of initial associated objects having a negative association relationship with the target account in the initial associated objects;

[0013] determining the associated object corresponding to the target account in the initial associated objects corresponding to the target account, wherein a number of initial associated objects having a positive association relationship with the target account in the associated object is less than or equal to a first number threshold, and a number of initial associated objects having a negative association relationship with the target account in the associated object is less than or equal to a second number threshold.

[0014] In a possible implementation, the attribute vector of the target account and the attribute vector of the associated object corresponding to the target account are obtained, including:

[0015] determining the attribute vector of the target account based on a preset vector corresponding to each attribute parameter of the target account;

[0016] determining the attribute vector of the associated object based on a preset vector corresponding to each attribute parameter of the associated object.

[0017] In a possible implementation, the feature extraction model includes a global feature extraction model, a local feature extraction model, and a combination model.

[0018] determining the feature vector of the target account and the feature vector of the associated object based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and the trained feature extraction model, comprises:

[0019] determining the global feature vector of the target account and the global feature vector of the associated object based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and the trained global feature extraction model;

[0020] determining the local feature vector of the target account and the local feature vector of the associated object based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and the trained local feature extraction model;

[0021] determining the feature vector of the target account and the feature vector of the associated object based on the global feature vector of the target account, the local feature vector of the target account, the global feature vector of the associated object, the local feature vector of the associated object, and the trained combination model.

[0022] In a possible implementation, the global feature extraction model comprises an initialization module, an updating module, and a result module;

[0023] determining the global feature vector of the target account and the global feature vector of the associated object based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and the trained global feature extraction model, comprises:

[0024] determining the initial weight vector corresponding to the target account and the initial weight vector corresponding to each associated object based on the attribute vector of the target account, the attribute vector of the associated object, n reference vectors, and the initialization module, wherein the initial weight vector comprises n initial weight values, and the n is a preset number;

[0025] updating the initial weight vector corresponding to the target account and the initial weight vector corresponding to each associated object based on the initial weight vector corresponding to the target account, the initial weight vector corresponding to each associated object, the association relationship between the associated object and the target account, and the updating module, to obtain the weight vector corresponding to the target account and the weight vector corresponding to each associated object, wherein the weight vector comprises n weight values;

[0026] determine a global feature vector of the target account and a global feature vector of the associated object based on the weight vector corresponding to the target account, the weight vector corresponding to each associated object, the n reference vectors, and the result module.

[0027] In a possible implementation, the initialization module includes an MLP (Multi-Layer Perceptron) sub-module.

[0028] The initial weight vector corresponding to the target account and the initial weight vector corresponding to the associated object are determined based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, n reference vectors, and the initialization module, including:

[0029] The attribute vector of the target account and the attribute vector of the associated object are respectively input into the MLP sub-module to obtain an adjustment vector of the target account and an adjustment vector of the associated object.

[0030] The adjustment vector of the target account and the adjustment vector of each associated object are respectively subjected to vector inner product processing with the n reference vectors to obtain n initial weight values corresponding to the target account and n initial weight values corresponding to each associated object as the initial weight vector corresponding to the target account and the initial weight vector corresponding to the associated object.

[0031] In a possible implementation, the initial weight vector corresponding to the target account and the initial weight vector corresponding to the associated object are respectively updated based on the initial weight vector corresponding to the target account, the initial weight vector corresponding to each associated object, the association relationship between the associated object and the target account, and the update module to obtain the weight vector corresponding to the target account and the weight vector corresponding to each associated object, including:

[0032] For each node in a node set composed of the target account and the associated object, the initial weight vectors corresponding to other nodes having a positive association relationship with the node are subjected to bit-by-bit addition processing to obtain a positive weight vector corresponding to the node, the initial weight vectors corresponding to other nodes having a negative association relationship with the node are subjected to bit-by-bit addition processing to obtain a negative weight vector corresponding to the node, and the initial weight vector corresponding to the node, the positive weight vector corresponding to the node, and the negative weight vector corresponding to the node are subjected to splicing processing to obtain an association weight vector corresponding to the node, wherein the node is the target account or the associated object.

[0033] The association weight vector corresponding to each node is respectively subjected to nonlinear transformation processing to obtain the weight vector corresponding to the target account and the weight vector corresponding to the association object.

[0034] In a possible implementation, the determination of the global feature vector of the target account and the global feature vector of the association object based on the weight vector corresponding to the target account, the weight vector corresponding to each association object, the n reference vectors, and the result module includes:

[0035] The n reference vectors are respectively subjected to weighting and processing using the weight vector corresponding to the target account and the weight vector corresponding to the association object to obtain the global feature vector of the target account and the global feature vector of the association object.

[0036] In a possible implementation, the determination of the local feature vector of the target account and the local feature vector of the association object based on the attribute vector of the target account, the attribute vector of the association object, the association relationship between the association object and the target account, and the trained local feature extraction model includes:

[0037] For each node in a node set composed of the target account and the association object, the attribute vectors of other nodes having a positive association relationship with the node are added to obtain a positive attribute vector of the node, the feature vectors of other nodes having a negative association relationship with the node are added to obtain a negative attribute vector of the node, and the attribute vector of the node, the positive attribute vector of the node, and the negative attribute vector of the node are spliced to obtain an association attribute vector of the node, wherein the node is the target account or the association object.

[0038] The association attribute vector of each node is respectively subjected to nonlinear transformation processing to obtain the local feature vector of the target account and the local feature vector of the association object.

[0039] In a possible implementation, the determination of the feature vector of the target account and the feature vector of the association object based on the global feature vector of the target account, the local feature vector of the target account, the global feature vector of the association object, the local feature vector of the association object, and the trained combination model includes:

[0040] For each node in a node set composed of the target account and the association object, the global feature vector of the node and the local feature vector of the node are spliced to obtain a spliced feature vector of the node, wherein the node is the target account or the association object.

[0041] The splicing feature vector of each node is respectively subjected to nonlinear transformation processing to obtain the feature vector of the target account and the feature vector of the associated object.

[0042] In a possible implementation, the determining of the target object displayed to the target account based on the currently stored feature vector of the target account and the feature vectors of the plurality of objects comprises:

[0043] The correlation degree between the feature vector of the target account and the feature vector of each object is respectively determined from the currently stored feature vector of the target account and the feature vectors of the plurality of objects.

[0044] The target object displayed to the target account is determined based on the correlation degree between the feature vector of the target account and the feature vector of each object.

[0045] In a possible implementation, the method further comprises:

[0046] The sample input data and the benchmark output data are obtained, wherein the sample input data is an attribute vector of a sample account, an attribute vector of a sample associated object corresponding to the sample account, and an association relationship between the sample associated object and the sample account, the sample associated object is an object having an association relationship with the sample account from a plurality of objects for display, and the benchmark output data comprises an association value corresponding to each sample associated object and the sample account, the association value corresponding to a sample associated object having a positive association relationship with the sample account is 1, and the association value corresponding to a sample associated object having a negative association relationship with the sample account is 0.

[0047] Based on the attribute vector of the sample account, the attribute vector of the sample associated object, n reference vectors to be trained, and an initialization module to be trained, an initial weight vector corresponding to the sample account and an initial weight vector corresponding to each sample associated object are determined.

[0048] Based on the initial weight vector corresponding to the sample account, the initial weight vector corresponding to each sample associated object, the association relationship between the sample associated object and the sample account, and an update module to be trained, a weight vector corresponding to the sample account and a weight vector corresponding to the sample associated object are determined.

[0049] Based on the weight vector corresponding to the sample account, the weight vector corresponding to each sample associated object, the n reference vectors to be trained, and a result module to be trained, a global feature vector of the sample account and a global feature vector of the sample associated object are determined.

[0050] determine the local feature vector of the sample account and the local feature vector of the sample associated object based on the attribute vector of the sample account, the attribute vector of the sample associated object, the association relationship between the sample associated object and the sample account, and the local feature extraction model to be trained;

[0051] determine the feature vector of the sample account and the feature vector of the sample associated object based on the global feature vector of the sample account, the local feature vector of the sample account, the global feature vector of the sample associated object, the local feature vector of the sample associated object, and the combination model to be trained;

[0052] calculate the association degree between the feature vector of the sample account and the feature vector of each sample associated object respectively to obtain actual output data;

[0053] train the n reference vectors to be trained, the initialization module to be trained, the update module to be trained, the result module to be trained, the local feature extraction model to be trained, and the combination model to be trained based on the actual output data and the reference output data, to obtain the n reference vectors trained, the initialization module trained, the update module trained, the result module trained, the local feature extraction model trained, and the combination model trained.

[0054] In a second aspect, a device for pushing display information is provided, and the device comprises:

[0055] an obtaining module, configured to obtain an attribute vector of a target account and an attribute vector of an associated object corresponding to the target account, wherein the associated object is an object having an association relationship with the target account in a plurality of objects for display, and the association relationship is a positive association relationship or a negative association relationship;

[0056] a determining module, configured to determine a feature vector of the target account and a feature vector of the associated object based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and a trained feature extraction model;

[0057] a display module, configured to, when a display triggering event of a target account is detected, determine a target object for display of the target account based on a currently stored feature vector of the target account and feature vectors of the plurality of objects, and push display information of the target object to the target account.

[0058] In a possible implementation manner, the obtaining module is further configured to:

[0059] In the plurality of objects for display, obtain objects having an association relationship with the target account within a preset time period, to obtain initial associated objects corresponding to the target account;

[0060] Determine a first number of initial associated objects having a positive association relationship with the target account and a second number of initial associated objects having a negative association relationship with the target account in the initial associated objects;

[0061] In the initial associated objects corresponding to the target account, determine the associated objects corresponding to the target account, wherein the number of initial associated objects having a positive association relationship with the target account in the associated objects is less than or equal to a first number threshold, and the number of initial associated objects having a negative association relationship with the target account in the associated objects is less than or equal to a second number threshold.

[0062] In a possible implementation, the obtaining module is configured to:

[0063] Determine an attribute vector of the target account based on a preset vector corresponding to each attribute parameter of the target account;

[0064] Determine an attribute vector of the associated object based on a preset vector corresponding to each attribute parameter of the associated object.

[0065] In a possible implementation, the feature extraction model includes a global feature extraction model, a local feature extraction model, and a combination model;

[0066] The determining module is configured to:

[0067] Determine a global feature vector of the target account and a global feature vector of the associated object based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and a trained global feature extraction model;

[0068] Determine a local feature vector of the target account and a local feature vector of the associated object based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and a trained local feature extraction model;

[0069] Determine a feature vector of the target account and a feature vector of the associated object based on the global feature vector of the target account, the local feature vector of the target account, the global feature vector of the associated object, the local feature vector of the associated object, and a trained combination model.

[0070] In a possible implementation, the global feature extraction model includes an initialization module, an update module, and a result module;

[0071] The determination module is configured to:

[0072] determine, based on the attribute vector of the target account, the attribute vector of the associated object, the n reference vectors, and the initialization module, an initial weight vector corresponding to the target account and an initial weight vector corresponding to each associated object, wherein the initial weight vector comprises n initial weight values, and n is a preset number;

[0073] update, based on the initial weight vector corresponding to the target account, the initial weight vector corresponding to each associated object, the association relationship between the associated object and the target account, and the update module, the initial weight vector corresponding to the target account and the initial weight vector corresponding to each associated object respectively, to obtain a weight vector corresponding to the target account and a weight vector corresponding to each associated object, wherein the weight vector comprises n weight values;

[0074] determine, based on the weight vector corresponding to the target account, the weight vector corresponding to each associated object, the n reference vectors, and the result module, a global feature vector of the target account and a global feature vector of the associated object.

[0075] In a possible implementation, the initialization module comprises an MLP sub-module.

[0076] The determination module is configured to:

[0077] input the attribute vector of the target account and the attribute vector of the associated object into the MLP sub-module respectively to obtain an adjustment vector of the target account and an adjustment vector of the associated object;

[0078] perform vector inner product processing on the adjustment vector of the target account and the adjustment vector of each associated object and the n reference vectors respectively to obtain n initial weight values corresponding to the target account and n initial weight values corresponding to each associated object as the initial weight vector corresponding to the target account and the initial weight vector corresponding to the associated object.

[0079] In a possible implementation, the determination module is configured to:

[0080] For each node in the node set composed of the target account and the association object, perform bit-by-bit addition processing on initial weight vectors corresponding to other nodes having positive association relationship with the node to obtain a positive weight vector corresponding to the node, perform bit-by-bit addition processing on initial weight vectors corresponding to other nodes having negative association relationship with the node to obtain a negative weight vector corresponding to the node, and perform splicing processing on the initial weight vector corresponding to the node, the positive weight vector corresponding to the node, and the negative weight vector corresponding to the node to obtain an association weight vector corresponding to the node, wherein the node is the target account or the association object;

[0081] Perform nonlinear transformation processing on the association weight vector corresponding to each node respectively to obtain a weight vector corresponding to the target account and a weight vector corresponding to the association object.

[0082] In a possible implementation, the determining module is configured to:

[0083] Perform weighted sum processing on the n reference vectors respectively using the weight vector corresponding to the target account and the weight vector corresponding to the association object to obtain a global feature vector of the target account and a global feature vector of the association object.

[0084] In a possible implementation, the determining module is configured to:

[0085] For each node in the node set composed of the target account and the association object, perform bit-by-bit addition processing on attribute vectors of other nodes having positive association relationship with the node to obtain a positive attribute vector of the node, perform bit-by-bit addition processing on attribute vectors of other nodes having negative association relationship with the node to obtain a negative attribute vector of the node, and perform splicing processing on the attribute vector of the node, the positive attribute vector of the node, and the negative attribute vector of the node to obtain an association attribute vector of the node, wherein the node is the target account or the association object;

[0086] Perform nonlinear transformation processing on the association attribute vector of each node respectively to obtain a local feature vector of the target account and a local feature vector of the association object.

[0087] In a possible implementation, the determining module is configured to:

[0088] For each node in the node set composed of the target account and the association object, perform splicing processing on the global feature vector of the node and the local feature vector of the node to obtain a splicing feature vector of the node, wherein the node is the target account or the association object;

[0089] The splicing feature vector of each node is respectively subjected to nonlinear transformation processing to obtain the feature vector of the target account and the feature vector of the associated object.

[0090] In a possible implementation, the display module is configured to:

[0091] Among the current stored feature vector of the target account and the feature vectors of the plurality of objects, the association degree between the feature vector of the target account and the feature vector of each object is determined respectively.

[0092] Based on the association degree between the feature vector of the target account and the feature vector of each object, the target object displayed to the target account is determined.

[0093] In a possible implementation, the device further includes a training module configured to:

[0094] Obtain sample input data and reference output data, wherein the sample input data is an attribute vector of a stored sample account, an attribute vector of a sample associated object corresponding to the sample account, and an association relationship between the sample associated object and the sample account, the sample associated object is an object having an association relationship with the sample account in a plurality of objects for display, the reference output data includes an association value corresponding to each sample associated object and the sample account, the association value corresponding to the sample associated object having a positive association relationship with the sample account is 1, and the association value corresponding to the sample associated object having a negative association relationship with the sample account is 0.

[0095] Based on the attribute vector of the sample account, the attribute vector of the sample associated object, n reference vectors to be trained, and an initialization module to be trained, determine an initial weight vector corresponding to the sample account and an initial weight vector corresponding to each sample associated object.

[0096] Based on the initial weight vector corresponding to the sample account, the initial weight vector corresponding to each sample associated object, the association relationship between the sample associated object and the sample account, and an update module to be trained, determine a weight vector corresponding to the sample account and a weight vector corresponding to the sample associated object.

[0097] Based on the weight vector corresponding to the sample account, the weight vector corresponding to each sample associated object, the n reference vectors to be trained, and a result module to be trained, determine a global feature vector of the sample account and a global feature vector of the sample associated object.

[0098] determine the local feature vector of the sample account and the local feature vector of the sample associated object based on the attribute vector of the sample account, the attribute vector of the sample associated object, the association relationship between the sample associated object and the sample account, and the local feature extraction model to be trained;

[0099] determine the feature vector of the sample account and the feature vector of the sample associated object based on the global feature vector of the sample account, the local feature vector of the sample account, the global feature vector of the sample associated object, the local feature vector of the sample associated object, and the combination model to be trained;

[0100] respectively calculate the association degree between the feature vector of the sample account and the feature vector of each sample associated object to obtain actual output data;

[0101] train the n reference vectors to be trained, the initialization module to be trained, the update module to be trained, the result module to be trained, the local feature extraction model to be trained, and the combination model to be trained based on the actual output data and the reference output data, to obtain trained n reference vectors, trained initialization module, trained update module, trained result module, trained local feature extraction model, and trained combination model.

[0102] In a third aspect, a computer device is provided, which includes a processor and a memory. The memory stores at least one instruction, which is loaded and executed by the processor to implement the operations performed by the method of pushing display information.

[0103] In a fourth aspect, a computer readable storage medium is provided, which stores at least one instruction. The instruction is loaded and executed by the processor to implement the operations performed by the method of pushing display information.

[0104] The technical scheme provided by the embodiments of the present application has the beneficial effects that: the scheme mentioned in the embodiments of the present application can determine the feature vector of the target account and the feature vector of each associated object based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and the feature extraction model trained, determine the feature vector of the plurality of accounts and the feature vector of the plurality of objects stored by using the above method, and then, when detecting the display triggering event of the target account, determine the target object displayed to the target account based on the feature vector of the target account and the feature vector of the plurality of objects currently stored, and then push the display information of the target object to the target account. In the above method, the determined associated object can include the object having a positive association relationship with the target account, and can also include the object having a negative association relationship with the target account, so that the feature vector of the target account determined is the feature vector that fuses the information of the objects liked and the objects disliked by the target account, and then, the target object determined by using the feature vector is also relatively accurate, the conversion rate of the pushed display information is improved, and the use experience of the user is improved. BRIEF DESCRIPTION OF DRAWINGS

[0105] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0106] Figure 1 is a method flowchart for pushing display information provided by the embodiments of the present application;

[0107] Figure 2 is a method flowchart for determining associated objects provided by the embodiments of the present application;

[0108] Figure 3 is a method flowchart for determining feature vectors provided by the embodiments of the present application;

[0109] Figure 4 is a method flowchart for determining feature vectors provided by the embodiments of the present application;

[0110] Figure 5 is a method flowchart for determining global feature vectors provided by the embodiments of the present application;

[0111] Figure 6 is a method flowchart for determining global feature vectors provided by the embodiments of the present application;

[0112] Figure 7 is a training method flowchart of a feature extraction model provided by the embodiments of the present application;

[0113] Figure 8 is a structural schematic diagram of a device for pushing display information provided by an embodiment of the present application;

[0114] Figure 9 is a structural block diagram of a server provided by an embodiment of the present application. DETAILED DESCRIPTION

[0115] In order to make the purpose, technical solutions and advantages of the present application clearer, the embodiments of the present application will be further described in detail below with reference to the drawings.

[0116] The present application provides a method for pushing display information, which can be implemented by a server. The server can be a single server or a server cluster composed of multiple servers.

[0117] The server can include a processor, a memory, a communication component, etc., and the processor is connected with the memory and the communication component respectively.

[0118] The processor can be a CPU (Central Processing Unit, central processor). The processor can be used for reading instructions and processing data, for example, obtaining the stored attribute vector of the target account, obtaining the stored attribute vector of the associated object corresponding to the target account, determining the feature vector of the target account and the feature vector of the associated object, determining the target object displayed to the target account, etc.

[0119] The memory can include a ROM (Read-Only Memory, read-only memory), a RAM (Random Access Memory, random access memory), a CD-ROM (Compact Disc Read-Only Memory, optical disc read-only memory), a magnetic disk, an optical data storage device, etc. The memory can be used for data storage, for example, storage of the attribute vector and the feature vector of the target account, storage of the attribute vector and the feature vector of the associated object, storage of intermediate data in the process of determining the feature vector of the target account and the feature vector of the associated object, storage of the data of the determined target object, storage of intermediate data in the process of determining the target object, etc.

[0120] The communication component can be a wired network connector, a WiFi (Wireless Fidelity, wireless fidelity) module, a Bluetooth module, a cellular communication module, etc. The communication component can be used for receiving and sending signals, for example, pushing the exhibition information of the target object to the target account, etc.

[0121] The server can be a background server of a shopping application, a search application, or the like, in which some relevant information or goods can be pushed to a user. For example, in a shopping application, some display information of goods can be pushed to a user; in a search application, some display information of goods or display information of news can be pushed to a user; in a friend-making application, display information of other users can be pushed to a user, and the like, which are not limited by the embodiments of the present application. Hereinafter, the method of pushing display information is described in detail by taking the example of pushing some display information of goods to a user in a shopping application.

[0122] It should be noted that the account (including the target account and other accounts) and the associated object (including but not limited to goods, news, accounts, and the like) in the embodiments of the present application can be collectively referred to as a node, that is, the node can be an account or an associated object.

[0123] Figure 1 is a flowchart of a method for pushing display information provided by the embodiments of the present application. Referring to Figure 1 , the embodiments include the following steps.

[0124] 101. Obtain an attribute vector of a target account and an attribute vector of an associated object corresponding to the target account.

[0125] The associated object is an object having an association relationship with the target account in a plurality of objects for display, and the association relationship is a positive association relationship or a negative association relationship.

[0126] In implementation, the attribute vector of each account and the attribute vector of each good are pre-stored in the database. When the feature vector of a node (including an account and a good) in the database is to be determined, a target account can be determined first, and then the attribute vector of the target account and the attribute vector of the associated object corresponding to the target account stored in the database are obtained.

[0127] The associated object is an object having a positive association relationship or a negative association relationship with the target account. For example, in a shopping application, some display information of goods is pushed to a user in a push interface, if the user clicks on the display information of A goods to enter the detail interface of A goods for browsing or purchases A goods, then A goods is the associated object of the account corresponding to the user, and A goods has a positive association relationship with the account corresponding to the user, if the user deletes the display information of B goods in the push interface or adds B goods to the blacklist, then B goods is the associated object of the account corresponding to the user, and B goods has a negative association relationship with the account corresponding to the user, if the user does not perform the above operation on other goods, then the other goods are not the associated object of the account corresponding to the user.

[0128] For example, in a search application, the user is pushed some news display information in the push interface, if the user clicks the display information of C news to browse it, the C news is the associated object of the account corresponding to the user, and the C news has a positive association relationship with the account corresponding to the user, if the user deletes the display information of D news in the push interface or adds D news to the blacklist, the D news is the associated object of the account corresponding to the user, and the D news has a negative association relationship with the account corresponding to the user, similarly, other news which does not perform the above operation is not the associated object of the account corresponding to the user.

[0129] For example, in a friend application, the user is pushed the account of other users who may be interested in the push interface, if the user browses the home page of other non-friend users (E users), the account of the E user can be determined as the associated object of the user, and the account of the E user has a positive association relationship with the account of the user, if the user adds the account of F user to the blacklist, the account of F user is the associated object of the account of the user, and the account of F user has a negative association relationship with the account of the user, similarly, the account of other users which does not perform the above operation is not the associated object of the account corresponding to the user.

[0130] The case that the object is an object other than a commodity, news and an account is similar to the above, which will not be repeated here.

[0131] Next, the determination method of the attribute vector of the target account and the attribute vector of the associated object will be described in detail, the corresponding processing can be: determining the attribute vector of the target account based on the preset vector corresponding to each attribute parameter of the target account. Determine the attribute vector of the associated object based on the preset vector corresponding to each attribute parameter of the associated object.

[0132] In implementation, the staff can pre-set the attribute types included in the attribute vector of the node, wherein each attribute type further includes at least one attribute parameter, and the staff can also pre-set the preset vector corresponding to each attribute parameter, and the preset vector corresponding to each attribute parameter is unique and not repeated. For at least one attribute parameter of the same attribute type, the length of the preset vector corresponding to the at least one attribute parameter is a constant value, for example, for the attribute type of "account ID", the attribute type includes multiple attribute parameters of account ID corresponding to each account, then the dimension of the preset vector corresponding to each account ID is the same, for example, it is a five-dimensional vector. Alternatively, the lengths of the preset vectors corresponding to the multiple attribute parameters included in the multiple attribute types are the same, which is not limited in the embodiments of the present application.

[0133] If the associated object is a commodity, the attribute type of the commodity can include at least one of the commodity type and the price range, if the associated object is news information, the attribute type of the news information can include the news type.

[0134] When it is necessary to determine the attribute vector of the target account, the preset vectors corresponding to the attribute parameters of each attribute type of the target account can be spliced in the order of the preset attribute types, and the preset vectors corresponding to the attribute parameters of the target account are spliced into a vector, which is the attribute vector of the target account. Similarly, when it is necessary to determine the attribute vector of the associated object, the preset vectors corresponding to the attribute parameters of each attribute type of the associated object can be spliced in the order of the preset attribute types of the goods, so as to obtain the attribute vector of the associated object.

[0135] Whenever a node is established, the server can determine the attribute vector of the node based on the above method and store it in the database. For example, when a new account is registered, the server can generate the attribute vector of the user account based on the attribute parameters of the account and store it in the database. For another example, when a merchant adds a new good, the server can generate the attribute vector of the good based on the attribute parameters of the good and store it in the database, and so on.

[0136] 102. Determine the feature vector of the target account and the feature vector of the associated object based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and the trained feature extraction model.

[0137] The association relationship between the associated object and the target account can be represented by a vector or a matrix, and each vector or matrix contains at least one element.

[0138] For the case where the associated object is a good or news, etc., since there is no association relationship between any two associated objects (for example, there is no association relationship between two goods), the association relationship only exists between the target account and the associated object, and therefore the association relationship between the associated object and the target account can be represented by a vector, and each element in the vector represents the association relationship between the associated object corresponding to the element and the target account.

[0139] For the case where the associated object is an account, etc., since any two associated objects can also have an association relationship (for example, two accounts can be friends or blacklisted), the association relationship between the associated object and the target account can be represented by a matrix, and the matrix can also include the association relationship between any two associated objects, and each element in the matrix represents the association relationship between the two nodes corresponding to the element.

[0140] In the above vector or matrix, if the association relationship between the nodes is a positive association relationship, the value of the element can be 1, if the association relationship is a negative association relationship, the value of the element can be 0, and for the case of using a matrix, if there is no association relationship between two associated objects, the value of the element can be 0.5 or other preset values, and the embodiments of the present application are not limited thereto. It can be understood that the association relationship of the node with itself can be considered as a positive association relationship, and the value of the corresponding element can be 1.

[0141] In implementation, the attribute vector of the target account, the attribute vector of the associated object, and the association relationship between the associated object and the target account can be input into the trained feature extraction model, so as to obtain the feature vector of the target account and the feature vector of each associated object, which can then be stored in the database. The obtained feature vectors of the nodes can be used to represent the preference information or type information of the nodes, because the information of other nodes having an association relationship with the node is fused.

[0142] The attribute vectors of a plurality of accounts and the attribute vectors of a plurality of associated objects corresponding to the accounts stored in the database can be periodically processed using the method in steps 101-102, so as to obtain the feature vectors of the accounts and the feature vectors of the associated objects that are more consistent with the current actual situation, so that the feature vectors can reflect the orientation and preference of the user corresponding to the account in the recent period, and the favorite or hated situation of each associated object.

[0143] 103、When the display triggering event of the target account is detected, the target object to be displayed to the target account is determined based on the currently stored feature vector of the target account and the feature vectors of a plurality of objects, and display information of the target object is pushed to the target account.

[0144] In implementation, when the server detects the display triggering event of the target account, the currently stored feature vector of the target account and the feature vectors of a plurality of objects can be obtained from the database, and based on the feature vectors, the target object to be pushed to the target account for display is determined, and then the display information of the target object is sent to the mobile terminal to which the target account belongs. After receiving the display information of the target object, the mobile terminal can display it in the recommendation interface for the user to view. It can be understood that when the server detects the display triggering event of any account, the target object to be displayed to the account can be determined based on the above method, and the display information of the target object is pushed to the account.

[0145] The above object can be a commodity, news, an account, etc. The above plurality of objects can be all the objects stored in the database, or a plurality of objects obtained after being filtered based on a preset condition, and the embodiments of the present application are not limited thereto.

[0146] Optionally, after the target object is determined, the multiple target objects can be sorted based on the attribute information and the historical operation statistical information of each target object, to obtain sorting information of the target objects, and then the target objects and the display information and the sorting information are sent to the mobile terminal to which the target account belongs. The mobile terminal can display the display information of the target objects based on the sorting information.

[0147] Next, the method for determining the target object is described in detail, and the corresponding processing can be as follows:

[0148] The correlation degree between the feature vector of the target account and the feature vector of each object is determined in the currently stored feature vector of the target account and the feature vectors of the multiple objects. Based on the correlation degree between the feature vector of the target account and the feature vector of each object, the target object displayed to the target account is determined.

[0149] In implementation, when the feature vector of the target account and the feature vectors of the multiple objects are obtained, the correlation degree between the feature vector of each object and the feature vector of the target account can be calculated respectively. There are multiple methods for calculating the correlation degree between two vectors, and one of them is provided as follows: the vector distance between the feature vector of the object and the feature vector of the target account can be calculated as the correlation degree, for example, the Euclidean distance, information entropy, and included angle cosine between two vectors can be calculated as the correlation degree between two vectors. The specific method is not limited in the embodiments of the present application.

[0150] After the correlation degree between the feature vector of each object and the feature vector of the target account is obtained, the target object displayed to the target account can be determined based on the correlation degree among the multiple objects. There are multiple possible methods for determining the target object, and several of them are provided as follows:

[0151] Firstly, the staff can pre-set a correlation degree threshold. After the correlation degree between the feature vector of each object and the feature vector of the target account is determined, the object corresponding to the correlation degree greater than the correlation degree threshold can be determined as the target object. The correlation degree threshold can be any reasonable value, for example, 0.5 or 0.6, and the like. The embodiments of the present application are not limited in this regard.

[0152] Secondly, the staff can pre-set an object number threshold. After the correlation degree between the feature vector of each object and the feature vector of the target account is determined, the objects can be arranged in descending order of the corresponding correlation degrees, and the first object number threshold objects can be determined as the target objects. The object number threshold can be any reasonable value, for example, 50 or 80, and the like. The embodiments of the present application are not limited in this regard.

[0153] In step 101, the feature vector of the associated object corresponding to the target account needs to be obtained, and before that, the associated object corresponding to the target account needs to be determined. The method for determining the associated object of the target account is described in detail as follows: Figure 2 The corresponding processing can be as follows:

[0154] 201. In the plurality of objects for display, the objects having an association relationship with the target account in a preset period are obtained, and the initial associated object corresponding to the target account is obtained.

[0155] The preset period can be any reasonable period, for example, it can be the last month or half a year, etc., and the embodiments of the present application do not limit this.

[0156] 202. Determine the first number of the initial associated object having a positive association relationship with the target account and the second number of the initial associated object having a negative association relationship with the target account in the initial associated object.

[0157] In implementation, in the determined initial associated object corresponding to the target account, there can be an associated object having a positive association relationship, and there can be an associated object having a negative association relationship. The number of the two kinds of associated objects is counted to obtain the number of the initial associated object having a positive association relationship with the target account, that is, the first number, and the number of the initial associated object having a negative association relationship with the target account, that is, the second number.

[0158] 203. In the initial associated object corresponding to the target account, the associated object corresponding to the target account is determined.

[0159] The number of the initial associated object having a positive association relationship with the target account in the associated object is less than or equal to the first number threshold, and the number of the initial associated object having a negative association relationship with the target account in the associated object is less than or equal to the second number threshold.

[0160] In implementation, the staff can pre-set the number threshold of the associated object having a positive association relationship with the target account in the associated object, that is, the first number threshold, and pre-set the number threshold of the associated object having a negative association relationship with the target account in the associated object, that is, the second number threshold.

[0161] Then, after the initial associated object corresponding to the target account is determined, a large number of initial associated objects can be obtained, at this time, the initial associated object corresponding to the target account can be screened based on the pre-set first number threshold and second number threshold.

[0162] If the first number is less than or equal to the first number threshold and the second number is less than or equal to the second number threshold in the initial association object, the initial association object corresponding to the target account can be directly determined as the association object corresponding to the target account.

[0163] If the first number is greater than the first number threshold and the second number is less than or equal to the second number threshold in the initial association object, the first number threshold initial association objects having a positive association relationship with the target account can be selected from all the initial association objects having a positive association relationship with the target account, and the first number threshold initial association objects and all the initial association objects having a negative association relationship with the target account are determined as the association object corresponding to the target account.

[0164] If the first number is less than or equal to the first number threshold and the second number is greater than the second number threshold in the initial association object, the second number threshold initial association objects having a negative association relationship with the target account can be selected from all the initial association objects having a negative association relationship with the target account, and the second number threshold initial association objects and all the initial association objects having a positive association relationship with the target account are determined as the association object corresponding to the target account.

[0165] If the first number is greater than the first number threshold and the second number is greater than the second number threshold in the initial association object, the first number threshold initial association objects having a positive association relationship with the target account can be selected from all the initial association objects having a positive association relationship with the target account, the second number threshold initial association objects having a negative association relationship with the target account can be selected from all the initial association objects having a negative association relationship with the target account, and the first number threshold initial association objects and the second number threshold initial association objects are determined as the association object corresponding to the target account.

[0166] Optionally, the selection operation in the above multiple cases can be random selection, or selecting the first number threshold or the second number threshold initial association objects closest to the current time, or any other reasonable method, and the embodiments of the present application are not limited thereto.

[0167] The feature extraction model in step 102 can include a global feature extraction model, a local feature extraction model, and a combination model. Based on the global feature extraction model, the global feature vector of each node can be obtained, based on the local feature extraction model, the local feature vector of each node can be obtained, and then based on the combination model, the global feature vector and the local feature vector of the node can be fused to obtain the feature vector of the node. Next, the processing in step 102 will be described in detail, as shown in Figure 3 and Figure 4 The corresponding processing can be as follows:

[0168] 301. Determine the global feature vector of the target account and the global feature vector of the associated object based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and the trained global feature extraction model.

[0169] In implementation, the attribute vector of the target account, the attribute vector of the associated object, and the association relationship between the associated object and the target account can be input into the trained global feature extraction model. For these nodes (i.e., the target account and the associated object), the global feature vector that integrates the information of other nodes having positive and / or negative association relationship with them is obtained. The global feature vector of the node can reflect the preference or type of the node.

[0170] In the embodiments of the present application, the global feature extraction model can include an initialization module, an update module, and a result module. The processing process of the global feature extraction model is described in detail as follows. Figure 5 and Figure 6 The corresponding processing can be as follows.

[0171] 301a. Determine the initial weight vector corresponding to the target account and the initial weight vector corresponding to each associated object based on the attribute vector of the target account, the attribute vector of the associated object, n reference vectors, and the initialization module.

[0172] The initial weight vector includes n initial weight values, and n is a preset number.

[0173] In implementation, before step 102, the feature extraction model needs to be trained first to obtain the trained feature extraction model, i.e., the global feature extraction model, the local feature extraction model, and the combination model are trained to obtain the trained global feature extraction model, the trained local feature extraction model, and the trained combination model. It can be understood that the data corresponding to the n reference vectors are the adjustment parameters in the global feature extraction model. In the process of training the global feature extraction model, the n reference vectors to be trained in the global feature extraction model are also trained to obtain the trained global feature extraction model and the trained n reference vectors therein. The reference vectors are described in more detail in the following content, and will not be described here.

[0174] After the attribute vector of the target account and the attribute vector of the associated object are determined, the attribute vector of the target account, the attribute vector of the associated object, and n reference vectors are input into the trained initialization module. The initialization module can output the initial weight vector corresponding to the target account and the initial weight vector corresponding to each associated object.

[0175] The initial weight vector corresponding to the node (the target account or the associated object) includes n initial weight values, and each initial weight value corresponds to a reference vector, and each initial weight value is used to represent the weight of the reference vector corresponding to the initial weight value in the updated feature vector of the node.

[0176] Next, the processing procedure in the initialization module is described in more detail, as follows:

[0177] The initialization module includes an MLP submodule. The attribute vector of the target account and the attribute vector of the associated object are respectively input into the MLP submodule to obtain the adjustment vector of the target account and the adjustment vector of the associated object. The adjustment vector of the target account and the adjustment vector of each associated object are respectively subjected to vector inner product processing with the n reference vectors to obtain n initial weight values corresponding to the target account and n initial weight values corresponding to each associated object as the initial weight vector corresponding to the target account and the initial weight vector corresponding to the associated object.

[0178] In implementation, the attribute vector of the target account and the attribute vector of the associated object can be input into the MLP submodule, and the MLP submodule can adjust the attribute vector of the target account and the attribute vector of the associated object to output the adjustment vector of the target account and the adjustment vector of the associated object. Then, for each node in the node set composed of the target account and the associated object, the adjustment vector of the node is subjected to vector inner product processing (also referred to as dot product processing) with each reference vector to obtain n initial weight values, and the initial weight values are arranged in the order of the corresponding reference vectors, so that the initial weight vector corresponding to the node can be obtained.

[0179] 301b, based on the initial weight vector corresponding to the target account, the initial weight vector corresponding to each associated object, the association relationship between the associated object and the target account, and the update module, the initial weight vector corresponding to the target account and the initial weight vector corresponding to the associated object are updated respectively to obtain the weight vector corresponding to the target account and the weight vector corresponding to each associated object.

[0180] The weight vector includes n weight values.

[0181] In this step, the initial weight vector corresponding to each node determined in the previous step (i.e., step 301a) is updated to obtain the weight vector corresponding to each node, so that each weight value in the weight vector corresponding to each node is more consistent with the actual value and more accurate than the initial weight value in the corresponding initial weight vector.

[0182] Next, the processing procedure in the update module is described in more detail, as follows:

[0183] For each node in the node set composed of the target account and the associated object, the initial weight vectors corresponding to other nodes having positive association relationship with the node are processed by bit-by-bit addition to obtain a positive weight vector corresponding to the node, the initial weight vectors corresponding to other nodes having negative association relationship with the node are processed by bit-by-bit addition to obtain a negative weight vector corresponding to the node, and the initial weight vector corresponding to the node, the positive weight vector corresponding to the node and the negative weight vector corresponding to the node are processed by splicing to obtain an association weight vector corresponding to the node. Wherein, the node is the target account or the associated object. The association weight vector corresponding to each node is processed by nonlinear transformation respectively to obtain a weight vector corresponding to the target account and a weight vector corresponding to the associated object.

[0184] In implementation, if there is an association relationship between each associated object, the association relationship between the associated object and the target account can be replaced by the association relationship between each node in the set, and if there is no association relationship between each associated object, the association relationship between the associated object and the target account can be directly used.

[0185] After the initial weight vector corresponding to each node, the association relationship between the associated object and the target account or the association relationship between each node is input into the updating module, the updating module will process each node as follows:

[0186] Based on the association relationship, other nodes having positive association relationship with the node and other nodes having negative association relationship with the node are determined, then the initial weight vectors corresponding to the other nodes having positive association relationship are processed by bit-by-bit addition to obtain a positive weight vector corresponding to the node, and the initial weight vectors corresponding to the other nodes having negative association relationship are processed by bit-by-bit addition to obtain a negative weight vector corresponding to the node. It can be understood that the bit-by-bit addition processing is to add the first element in each vector to obtain the first element of the new vector, add the second element in each vector to obtain the second element of the new vector, and so on, so as to obtain each element in the new vector. Here, the new vector can refer to the positive weight vector or the negative weight vector.

[0187] After obtaining the positive weight vector corresponding to the node and the negative weight vector corresponding to the node, the initial weight vector corresponding to the node, the positive weight vector corresponding to the node and the negative weight vector corresponding to the node can be processed by splicing, i.e. the head and tail of each vector are connected to form a vector, so as to obtain an association weight vector corresponding to the node.

[0188] By the above method, the associated weight vector corresponding to each node in the node set is obtained, and then the associated weight vector corresponding to each node is subjected to nonlinear transformation processing, the dimension of the associated weight vector of the node is adjusted to n dimensions, so that the weight vector corresponding to each node is obtained, and in the weight vector, n weight values are included. Each weight value is used to represent the weight of the reference vector corresponding to the weight value in the updated feature vector of the node, which is an update of the initial weight value corresponding to the weight value, so that the obtained weight value is more accurate.

[0189] For example, the node set includes A node, B node, C node, D node and E node, and three reference vectors are set. The other nodes having a positive association relationship with the A node are B node and C node, and the other nodes having a negative association relationship with the A node are D node and E node. The initial weight vector corresponding to the A node is (a1, a2, a3), the initial weight vector corresponding to the B node is (b1, b2, b3), the initial weight vector corresponding to the C node is (c1, c2, c3), the initial weight vector corresponding to the D node is (d1, d2, d3), and the initial weight vector corresponding to the E node is (e1, e2, e3). The obtained positive weight vector corresponding to the A node is (b1+c1, b2+c2, b3+c3), the negative weight vector corresponding to the A node is (d1+e1, d2+e2, d3+e3), and the associated weight vector corresponding to the A node is (a1, a2, a3, b1+c1, b2+c2, b3+c3, d1+e1, d2+e2, d3+e3). Then, the associated weight vector corresponding to the A node is subjected to nonlinear transformation processing, and the weight vector corresponding to the A node is obtained. The weight vector is a three-dimensional vector, including three weight values, and each weight value is used to represent the weight of the reference vector corresponding to the weight value in the updated feature vector corresponding to the node A.

[0190] Optionally, after the initial weight vector of each node in the node set is updated once by using the above method, the obtained weight vector corresponding to each node can also be updated by using the above method, to obtain the weight vector corresponding to the node after the second update. The staff can set the threshold value of the number of cyclic updates, and update the initial weight vector of each node in the node set multiple times by using the above method, to obtain the weight vector corresponding to each node after the update. The specific value of the set threshold value of the number of cyclic updates is not limited in the embodiments of the application, which can be 1, 2 or more times, etc.

[0191] 301c, determining the global feature vector of the target account and the global feature vector of the associated object based on the weight vector corresponding to the target account, the weight vector corresponding to each associated object, the n reference vectors and the result module.

[0192] In implementation, after obtaining the weight vector corresponding to each node, the global feature vector of each node can be determined based on each weight value in the weight vector and the reference vector.

[0193] Next, the processing procedure inside the result module is described in more detail, which corresponds to the following:

[0194] The n reference vectors are weighted and processed using the weight vector corresponding to the target account and the weight vector corresponding to the associated object, to obtain the global feature vector of the target account and the global feature vector of the associated object.

[0195] In implementation, each node in the node set is processed as follows: based on the n weight values in the weight vector corresponding to the node, the n reference vectors are weighted and processed, thereby obtaining the global feature vector of the node, that is, for the weight vector corresponding to a node, the n weight values included in the weight vector are multiplied by the corresponding reference vectors respectively, and the reference vectors after multiplication processing are added bit by bit, thereby obtaining the vector after the n reference vectors are weighted and processed, which is the global feature vector of the node.

[0196] For example, there are three reference vectors, which are (X1, X2, X3, X4), (Y1, Y2, Y3, Y4) and (Z1, Z2, Z3, Z4), and if the weight vector corresponding to node F is (f1, f2, f3), then the global feature vector of the node is (f1X1+f2Y1+f3Z1, f1X2+f2Y2+f3Z2, f1X3+f2Y3+f3Z3, f1X4+f2Y4+f3Z4).

[0197] It can be understood that in step 301, the n reference vectors can be used to represent the type vectors of the plurality of nodes stored in the database, that is, the n reference vectors divide the plurality of nodes stored in the database into n types. Then, the n weight values included in the weight vector corresponding to each node determined in the global feature extraction model can be used to represent the weight of the node belonging to each type.

[0198] 302、based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and the trained local feature extraction model, determine the local feature vector of the target account and the local feature vector of the associated object.

[0199] In implementation, the global feature vector of each node in the node set determined in step 301 can be used to represent which type the node is more inclined to, but there can also be nodes with special orientation in the database, for example, taking account number and goods as an example, the user corresponding to B account dislikes various types of goods and does not show the goods he likes. This kind of user is relatively small, so a local feature vector can be set to represent the preference of the node.

[0200] In implementation, the attribute vector of the target account, the attribute vector of the associated object, and the association relationship between the associated object and the target account can be input into the trained local feature extraction model to obtain the output local feature vector of the target account and the local feature vector of each associated object.

[0201] Similarly, if there is an association relationship between each associated object, the association relationship of multiple nodes in the node set can be used to replace the association relationship between the associated object and the target account, so that the feature vector of the target account, the feature vector of the associated object, and the association relationship of each node are input into the trained local feature extraction model to obtain the local feature vector of each node.

[0202] Next, the processing process of the local feature extraction model is described in detail. The corresponding processing can be as follows:

[0203] For each node in the node set composed of the target account and the associated object, the attribute vectors of other nodes having positive association relationship with the node are added to obtain the positive attribute vector of the node, the attribute vectors of other nodes having negative association relationship with the node are added to obtain the negative attribute vector of the node, and the attribute vector of the node, the positive attribute vector of the node and the negative attribute vector of the node are spliced to obtain the associated attribute vector of the node, wherein the node is the target account or the associated object. The associated attribute vector of each node is respectively subjected to nonlinear transformation processing to obtain the local feature vector of the target account and the local feature vector of the associated object.

[0204] In implementation, each node in the node set is respectively processed as follows:

[0205] Based on the correlation in the input local feature extraction model, other nodes having positive correlation with the node and other nodes having negative correlation with the node are determined, the attribute vectors of the other nodes having positive correlation are processed by bit-by-bit addition, and the positive attribute vector of the node is obtained. Similarly, the attribute vectors of the other nodes having negative correlation are processed by bit-by-bit addition, and the negative attribute vector of the node is obtained. Then, the attribute vector of the node, the positive attribute vector of the node and the negative attribute vector of the node are spliced, that is, the three vectors are combined into one vector, and the correlation attribute vector of the node is obtained.

[0206] The correlation attribute vector of each node in the node set is obtained using the above method, and then the correlation attribute vector of each node is processed by nonlinear transformation, thereby obtaining the local feature vector of each node.

[0207] 303、Based on the global feature vector of the target account, the local feature vector of the target account, the global feature vector of the associated object, the local feature vector of the associated object and the trained combination model, the feature vector of the target account and the feature vector of the associated object are determined.

[0208] In implementation, after obtaining the global feature vector and the local feature vector of each node in the node set, the global feature vector and the local feature vector of each node can be input into the trained combination model for fusion, thereby obtaining the feature vector of each node.

[0209] Next, the processing process of the combination model is described in detail, and the corresponding processing can be as follows:

[0210] For each node in the node set composed of the target account and the associated object, the global feature vector of the node and the local feature vector of the node are spliced to obtain the spliced feature vector of the node, wherein the node is the target account or the associated object. The spliced feature vector of each node is processed by nonlinear transformation to obtain the feature vector of the target account and the feature vector of the associated object.

[0211] In implementation, for each node in the node set, the following processing is performed: the global feature vector and the local feature vector of the node are spliced, that is, the two vectors are combined into one vector, which is the spliced feature vector of the node.

[0212] The spliced feature vector of each node in the node set is obtained using the above method, and then the spliced feature vector of each node is processed by nonlinear transformation, thereby obtaining the feature vector of each node.

[0213] Next, the training process of the above feature extraction model is described in more detail, as follows: Figure 7As shown, the corresponding processing can be as follows:

[0214] 701. Obtain sample input data and reference output data.

[0215] The sample input data is a stored attribute vector of a sample account, an attribute vector of a sample associated object corresponding to the sample account, and an association relationship between the sample associated object and the sample account. The sample associated object is an object having an association relationship with the sample account in a plurality of objects for display. The reference output data includes an association value corresponding to each sample associated object and the sample account. The association value corresponding to the sample associated object having a positive association relationship with the sample account is 1, and the association value corresponding to the sample associated object having a negative association relationship with the sample account is 0.

[0216] In implementation, a sample set can be obtained first, and the sample set includes a plurality of sample data. Each sample data includes sample input data and reference output data. When the feature extraction model needs to be trained, a sample data in the sample set can be obtained first, i.e., the sample input data and the reference output data are obtained.

[0217] 702. Determine an initial weight vector corresponding to the sample account and an initial weight vector corresponding to each sample associated object based on the attribute vector of the sample account, the attribute vector of the sample associated object, n reference vectors to be trained, and an initialization module to be trained.

[0218] In implementation, when training is performed, the attribute vector of the sample account, the attribute vector of the associated object, and the n reference vectors to be trained can be input into the initialization module to be trained, to obtain the output initial weight vector corresponding to the sample account and the initial weight vector corresponding to each sample associated object.

[0219] The n reference vectors can be n vectors composed of parameters in the global feature extraction model. When the global feature extraction model is run for the first time (i.e., when the global feature extraction model is trained for the first time), the values of the elements in the n reference vectors can be randomly selected. In the subsequent training (also referred to as parameter adjustment) of the global feature extraction model, the values of the elements in the n reference vectors are adjusted, thereby obtaining the trained n reference vectors.

[0220] 703. Determine a weight vector corresponding to the sample account and a weight vector corresponding to each sample associated object based on the initial weight vector corresponding to the sample account, the initial weight vector corresponding to each sample associated object, the association relationship between the sample associated object and the sample account, and an update module to be trained.

[0221] In implementation, the initial weight vector corresponding to the sample account, the initial weight vector corresponding to each sample associated object, and the association relationship between the sample associated object and the sample account are input into the to-be-trained update module to obtain the output weight vector corresponding to the sample account and the weight vector corresponding to the sample associated object.

[0222] 704、Based on the weight vector corresponding to the sample account, the weight vector corresponding to each sample associated object, the n reference vectors to be trained, and the result module to be trained, the global feature vector of the sample account and the global feature vector of the sample associated object are determined.

[0223] In implementation, the initial weight vector corresponding to the sample account, the initial weight vector corresponding to each sample associated object, and the association relationship between the sample associated object and the sample account are input into the to-be-trained update module to obtain the output weight vector corresponding to the sample account and the weight vector corresponding to the sample associated object.

[0224] 705、Based on the attribute vector of the sample account, the attribute vector of the sample associated object, the association relationship between the sample associated object and the sample account, and the local feature extraction model to be trained, the local feature vector of the sample account and the local feature vector of the sample associated object are determined.

[0225] In implementation, the attribute vector of the sample account, the attribute vector of each sample associated object, and the association relationship between the sample associated object and the sample account are input into the to-be-trained local feature extraction model to obtain the output local feature vector of the sample account and the local feature vector of each sample associated object.

[0226] 706、Based on the global feature vector of the sample account, the local feature vector of the sample account, the global feature vector of the sample associated object, the local feature vector of the sample associated object, and the to-be-trained combination model, the feature vector of the sample account and the feature vector of the sample associated object are determined.

[0227] In implementation, after obtaining the global feature vector and the local feature vector of the sample account, the global feature vector and the local feature vector of each sample associated object, they are input into the to-be-trained combination model to obtain the output feature vector of the sample account and the feature vector of each sample associated object.

[0228] 707、Calculate the association degree between the feature vector of the sample account and the feature vector of each sample associated object respectively to obtain the actual output data.

[0229] In implementation, after obtaining the feature vector of the sample account and the feature vector of each sample associated object, the correlation degree between the feature vector of each sample associated object and the feature vector of the sample account can be calculated, and a value in the range of [0, 1] is obtained, that is, the actual output data. It can be understood that the data form of the actual output data can be a vector, and each element of the vector represents the correlation value between the sample associated object corresponding to the element and the sample account.

[0230] It can be understood that there can be multiple methods for calculating the correlation degree between the feature vector of the sample associated object and the feature vector of the sample account, and two methods are described below:

[0231] Firstly, the vector distance between the feature vector of the sample associated object and the feature vector of the sample account can be calculated, for example, the Euclidean distance, information entropy, and cosine of the angle between the two vectors can be calculated as the correlation degree between the two vectors, and the specific method is not limited in the embodiments of the present application.

[0232] Secondly, the feature vector of each sample associated object can be processed as follows: the feature vector of the sample associated object and the feature vector of the sample account can be input into a fully connected layer to obtain a value, which is determined as the correlation degree between the feature vector of the sample associated object and the feature vector of the sample account.

[0233] The embodiments of the present application can use any of the above methods to calculate the correlation degree between the feature vector of the sample associated object and the feature vector of the sample account, or other reasonable methods, and the embodiments of the present application are not limited thereto.

[0234] 708、Based on the actual output data and the benchmark output data, the n reference vectors to be trained, the initialization module to be trained, the update module to be trained, the result module to be trained, the local feature extraction model to be trained, and the combination model to be trained are trained to obtain the n reference vectors trained, the initialization module trained, the update module trained, the result module trained, the local feature extraction model trained, and the combination model trained.

[0235] In implementation, the actual output data and the benchmark output data can be input into a loss function to obtain a loss value, and then based on the loss value, the n reference vectors to be trained, the initialization vector to be trained, the update module to be trained, the result module to be trained, the local feature extraction model to be trained, and the combination model to be trained are trained (which can also be referred to as parameter tuning).

[0236] After one training, another sample data can be obtained from the sample set, and the n reference vectors to be trained, the initialization vector to be trained, the update module to be trained, the result module to be trained, the local feature extraction model to be trained, and the combination model to be trained are trained again. The network is trained repeatedly by using multiple different sample data, and when a preset ending condition is reached, the training can be ended, and the global feature extraction model (including n reference vectors, an initialization module, an update module, and a result module) obtained after the last training, the local feature extraction model, and the combination model are determined as a trained global feature extraction model, a trained local feature extraction model, and a trained combination model.

[0237] The preset ending condition can be set in multiple ways, and the following are three of them:

[0238] The first kind is that the staff can pre-set a threshold of the number of training, and when the actual number of training (i.e., the number of adjusting parameters) reaches the threshold, the training can be stopped. For example, the threshold can be set to 200 or 300, or other reasonable number, which is not limited in the embodiments of the present application.

[0239] The second kind is that the staff can pre-set a threshold of loss value, and when the loss value of the last preset number of times is less than the threshold, the training can be stopped. For example, the preset number of times can be set to 3, and the threshold of loss value can be set to 0.1, and when the loss value calculated in the last three training is less than 0.1, the training can be stopped.

[0240] The third kind is that when the actual number of training reaches the preset threshold, and the loss value of the last preset number of times is less than the threshold, the training can be stopped.

[0241] All the optional technical solutions described above can be combined to form optional embodiments of the present application, which will not be repeated here.

[0242] The scheme mentioned in the embodiments of the present application can determine the feature vector of the target account and the feature vector of each associated object based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and the feature extraction model trained. The feature vector of the plurality of accounts stored and the feature vector of the plurality of objects are determined using the above method. Then, when the display triggering event of the target account is detected, the target object displayed to the target account can be determined based on the currently stored feature vector of the target account and the feature vector of the plurality of objects, and then the display information of the target object can be pushed to the target account. In the above method, the determined associated object can include both the object having a positive association relationship with the target account and the object having a negative association relationship with the target account. Therefore, the feature vector of the target account determined is the feature vector that integrates the information of the objects liked and the objects disliked by the target account. Then, the target object determined using the feature vector is also relatively accurate, which improves the conversion rate of the pushed display information and improves the user experience.

[0243] The embodiments of the present application provide a device for pushing display information. The device can be a computer device in the above embodiments, as shown in Figure 8 The device comprises:

[0244] The acquisition module 810 is configured to acquire the attribute vector of the target account and the attribute vector of the associated object corresponding to the target account. The associated object is an object having an association relationship with the target account in the plurality of objects for display. The association relationship is a positive association relationship or a negative association relationship.

[0245] The determination module 820 is configured to determine the feature vector of the target account and the feature vector of the associated object based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and the trained feature extraction model.

[0246] The display module 830 is configured to determine the target object displayed to the target account based on the currently stored feature vector of the target account and the feature vector of the plurality of objects when the display triggering event of the target account is detected, and push the display information of the target object to the target account.

[0247] In a possible implementation, the acquisition module 810 is further configured to:

[0248] In the plurality of objects for display, the associated object having an association relationship with the target account in a preset time period is acquired to obtain the initial associated object corresponding to the target account.

[0249] determine a first number of initial association objects having a positive association relationship with the target account and a second number of initial association objects having a negative association relationship with the target account in the initial association objects;

[0250] determine the association object corresponding to the target account in the initial association objects corresponding to the target account, wherein a number of initial association objects having a positive association relationship with the target account in the association object is less than or equal to a first number threshold, and a number of initial association objects having a negative association relationship with the target account in the association object is less than or equal to a second number threshold.

[0251] In a possible implementation, the obtaining module 810 is configured to:

[0252] determine an attribute vector of the target account based on a preset vector corresponding to each attribute parameter of the target account;

[0253] determine an attribute vector of the association object based on a preset vector corresponding to each attribute parameter of the association object.

[0254] In a possible implementation, the feature extraction model includes a global feature extraction model, a local feature extraction model, and a combination model.

[0255] The determining module 820 is configured to:

[0256] determine a global feature vector of the target account and a global feature vector of the association object based on the attribute vector of the target account, the attribute vector of the association object, the association relationship between the association object and the target account, and the trained global feature extraction model;

[0257] determine a local feature vector of the target account and a local feature vector of the association object based on the attribute vector of the target account, the attribute vector of the association object, the association relationship between the association object and the target account, and the trained local feature extraction model;

[0258] determine a feature vector of the target account and a feature vector of the association object based on the global feature vector of the target account, the local feature vector of the target account, the global feature vector of the association object, the local feature vector of the association object, and the trained combination model.

[0259] In a possible implementation, the global feature extraction model includes an initialization module, an updating module, and a result module.

[0260] The determining module 820 is configured to:

[0261] determine, based on the attribute vector of the target account, the attribute vector of the associated object, the n reference vectors, and the initialization module, an initial weight vector corresponding to the target account and an initial weight vector corresponding to each associated object, wherein the initial weight vector includes n initial weight values, and n is a preset number;

[0262] update, based on the initial weight vector corresponding to the target account, the initial weight vector corresponding to each associated object, the association relationship between the associated object and the target account, and the update module, the initial weight vector corresponding to the target account and the initial weight vector corresponding to each associated object respectively, to obtain a weight vector corresponding to the target account and a weight vector corresponding to each associated object, wherein the weight vector includes n weight values;

[0263] determine, based on the weight vector corresponding to the target account, the weight vector corresponding to each associated object, the n reference vectors, and the result module, a global feature vector of the target account and a global feature vector of the associated object.

[0264] In a possible implementation, the initialization module includes an MLP sub-module.

[0265] The determination module 820 is configured to:

[0266] input the attribute vector of the target account and the attribute vector of the associated object into the MLP sub-module respectively to obtain an adjustment vector of the target account and an adjustment vector of the associated object;

[0267] perform vector inner product processing on the adjustment vector of the target account and the adjustment vector of each associated object and the n reference vectors respectively to obtain n initial weight values corresponding to the target account and n initial weight values corresponding to each associated object as the initial weight vector corresponding to the target account and the initial weight vector corresponding to the associated object.

[0268] In a possible implementation, the determination module 820 is configured to:

[0269] for each node in a node set composed of the target account and the associated object, perform bit-by-bit addition processing on the initial weight vectors corresponding to other nodes having a positive association relationship with the node to obtain a positive weight vector corresponding to the node, perform bit-by-bit addition processing on the initial weight vectors corresponding to other nodes having a negative association relationship with the node to obtain a negative weight vector corresponding to the node, and perform splicing processing on the initial weight vector corresponding to the node, the positive weight vector corresponding to the node, and the negative weight vector corresponding to the node to obtain an association weight vector corresponding to the node, wherein the node is the target account or the associated object.

[0270] respectively, to obtain the weight vector corresponding to the target account and the weight vector corresponding to the associated object.

[0271] In a possible implementation, the determining module 820 is configured to:

[0272] respectively, to obtain the global feature vector of the target account and the global feature vector of the associated object.

[0273] In a possible implementation, the determining module 820 is configured to:

[0274] For each node in the node set composed of the target account and the associated object, the attribute vectors of other nodes having a positive association relationship with the node are added to obtain a positive attribute vector of the node, the feature vectors of other nodes having a negative association relationship with the node are added to obtain a negative attribute vector of the node, and the attribute vector of the node, the positive attribute vector of the node, and the negative attribute vector of the node are spliced to obtain an associated attribute vector of the node, wherein the node is the target account or the associated object.

[0275] respectively, to obtain the local feature vector of the target account and the local feature vector of the associated object.

[0276] In a possible implementation, the determining module 820 is configured to:

[0277] For each node in the node set composed of the target account and the associated object, the global feature vector of the node and the local feature vector of the node are spliced to obtain a spliced feature vector of the node, wherein the node is the target account or the associated object.

[0278] respectively, to obtain the feature vector of the target account and the feature vector of the associated object.

[0279] In a possible implementation, the displaying module 830 is configured to:

[0280] In the current stored feature vector of the target account and the feature vectors of the plurality of objects, an association degree between the feature vector of the target account and the feature vector of each object is determined.

[0281] determine the target object to be displayed to the target account based on the correlation between the feature vector of the target account and the feature vector of each object.

[0282] In a possible implementation, the apparatus further includes a training module configured to:

[0283] obtain sample input data and reference output data, wherein the sample input data is an attribute vector of a sample account, an attribute vector of a sample associated object corresponding to the sample account, and an association relationship between the sample associated object and the sample account, the sample associated object is an object having an association relationship with the sample account in a plurality of objects for display, the reference output data includes an association value corresponding to each sample associated object and the sample account, the association value corresponding to the sample associated object having a positive association relationship with the sample account is 1, and the association value corresponding to the sample associated object having a negative association relationship with the sample account is 0;

[0284] determine an initial weight vector corresponding to the sample account and an initial weight vector corresponding to each sample associated object based on the attribute vector of the sample account, the attribute vector of the sample associated object, n reference vectors to be trained, and an initialization module to be trained;

[0285] determine a weight vector corresponding to the sample account and a weight vector corresponding to the sample associated object based on the initial weight vector corresponding to the sample account, the initial weight vector corresponding to each sample associated object, the association relationship between the sample associated object and the sample account, and an update module to be trained;

[0286] determine a global feature vector of the sample account and a global feature vector of the sample associated object based on the weight vector corresponding to the sample account, the weight vector corresponding to each sample associated object, the n reference vectors to be trained, and a result module to be trained;

[0287] determine a local feature vector of the sample account and a local feature vector of the sample associated object based on the attribute vector of the sample account, the attribute vector of the sample associated object, the association relationship between the sample associated object and the sample account, and a local feature extraction model to be trained;

[0288] determine a feature vector of the sample account and a feature vector of the sample associated object based on the global feature vector of the sample account, the local feature vector of the sample account, the global feature vector of the sample associated object, the local feature vector of the sample associated object, and a combination model to be trained;

[0289] Calculate the association degree between the feature vector of the sample account and the feature vector of each sample associated object respectively to obtain actual output data;

[0290] Based on the actual output data and the reference output data, train the n reference vectors to be trained, the initialization module to be trained, the update module to be trained, the result module to be trained, the local feature extraction model to be trained, and the combined model to be trained to obtain the n reference vectors trained, the initialization module trained, the update module trained, the result module trained, the local feature extraction model trained, and the combined model trained.

[0291] It should be noted that the device for pushing display information provided in the above embodiments only divides the above functions by way of example, and in actual application, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device for pushing display information and the method for pushing display information provided in the above embodiments belong to the same concept, and the specific implementation process is detailed in the method embodiments, which will not be described here.

[0292] Figure 9 is a structural schematic diagram of a server provided by an embodiment of the application. The server 900 can have great differences due to different configurations or performances, and can include one or more central processors 901 and one or more memories 902. The memory 902 stores at least one instruction, which is loaded and executed by the processor 901 to implement the method provided by each method embodiment described above. Of course, the server can also have a wired or wireless network interface, a keyboard, an input and output interface, and other components for realizing the functions of the device, and will not be described here.

[0293] In the exemplary embodiments, a computer readable storage medium, such as a memory including instructions, is also provided, which can be executed by a processor in a terminal to complete the method for pushing display information in the above embodiments. The computer readable storage medium can be non-transitory. For example, the computer readable storage medium can be a ROM (read-only memory), a RAM (random access memory), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.

[0294] Those skilled in the art can understand that all or part of the steps of the above-mentioned embodiments can be completed by hardware, or can be instructed by a program to complete the related hardware, and the program can be stored in a computer readable storage medium. The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0295] It should be noted that the information (including but not limited to user equipment information, user personal information, etc.), data (including but not limited to data for analysis, stored data, displayed data, etc.) and signals (including but not limited to signals transmitted between user terminals and other devices, etc.) involved in the present application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards of relevant countries and regions. For example, the "account ID" involved in the present application is obtained under sufficient authorization.

[0296] The above only describes optional embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for pushing and displaying information, characterized in that, The method includes: Obtain the attribute vector of the target account and the attribute vector of the associated object corresponding to the target account, wherein the associated object is an object that has an association relationship with the target account among multiple objects used for display, and the association relationship is a positive association relationship or a negative association relationship; Based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and the trained feature extraction model, the feature vector of the target account and the feature vector of the associated object are determined. When a display trigger event for a target account is detected, based on the currently stored feature vector of the target account and the feature vectors of the multiple objects, the target object to be displayed to the target account is determined, and the display information of the target object is pushed to the target account. The feature extraction model includes a global feature extraction model, a local feature extraction model, and a combined model; The step of determining the feature vector of the target account and the feature vector of the associated object based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and the trained feature extraction model includes: Based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and the trained global feature extraction model, the global feature vector of the target account and the global feature vector of the associated object are determined. Based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and the trained local feature extraction model, the local feature vector of the target account and the local feature vector of the associated object are determined. Based on the global feature vector of the target account, the local feature vector of the target account, the global feature vector of the associated object, the local feature vector of the associated object, and the trained combined model, the feature vector of the target account and the feature vector of the associated object are determined.

2. The method according to claim 1, characterized in that, Before obtaining the attribute vector of the target account and the attribute vector of the associated object corresponding to the target account, the method further includes: Among the multiple objects used for display, the objects that are associated with the target account within a preset time period are obtained to obtain the initial associated object corresponding to the target account; Determine a first number of initial associated objects that have a positive association with the target account and a second number of initial associated objects that have a negative association with the target account; Among the initial associated objects corresponding to the target account, the associated objects corresponding to the target account are determined, wherein the number of initial associated objects with a positive association relationship with the target account is less than or equal to a first number threshold, and the number of initial associated objects with a negative association relationship with the target account is less than or equal to a second number threshold.

3. The method according to claim 1, characterized in that, The process of obtaining the attribute vector of the target account and the attribute vector of the associated object corresponding to the target account includes: The attribute vector of the target account is determined based on the preset vector corresponding to each attribute parameter of the target account; The attribute vector of the associated object is determined based on the preset vector corresponding to each attribute parameter of the associated object.

4. The method according to claim 1, characterized in that, The global feature extraction model includes an initialization module, an update module, and a result module; The step of determining the global feature vector of the target account and the global feature vector of the associated object based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and a trained global feature extraction model includes: Based on the attribute vector of the target account, the attribute vector of the associated object, n reference vectors and the initialization module, the initial weight vector corresponding to the target account and the initial weight vector corresponding to each associated object are determined, wherein the initial weight vector includes n initial weight values, and n is a preset number; Based on the initial weight vector corresponding to the target account, the initial weight vector corresponding to each associated object, the association relationship between the associated object and the target account, and the update module, the initial weight vector corresponding to the target account and the initial weight vector corresponding to the associated object are updated respectively to obtain the weight vector corresponding to the target account and the weight vector corresponding to each associated object, wherein the weight vector includes n weight values; Based on the weight vector corresponding to the target account, the weight vector corresponding to each associated object, the n reference vectors, and the result module, the global feature vector of the target account and the global feature vector of the associated object are determined.

5. The method according to claim 4, characterized in that, The initialization module includes a multilayer perceptron (MLP) submodule; the step of determining the initial weight vector corresponding to the target account and the initial weight vector corresponding to the associated object based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, n reference vectors, and the initialization module includes: The attribute vector of the target account and the attribute vector of the associated object are respectively input into the MLP submodule to obtain the adjustment vector of the target account and the adjustment vector of the associated object; The adjustment vector of the target account and the adjustment vector of each associated object are respectively processed by the dot product of the n reference vectors to obtain the n initial weight values ​​corresponding to the target account and the n initial weight values ​​corresponding to each associated object, which are used as the initial weight vector corresponding to the target account and the initial weight vector corresponding to the associated object.

6. The method according to claim 4, characterized in that, The step of updating the initial weight vector corresponding to the target account and the initial weight vector corresponding to each associated object based on the initial weight vector corresponding to the target account, the initial weight vector corresponding to each associated object, the association relationship between the associated object and the target account, and the update module, respectively, to obtain the weight vector corresponding to the target account and the weight vector corresponding to each associated object, includes: for each node in the node set composed of the target account and the associated object, performing positional addition on the initial weight vectors corresponding to other nodes with positive association relationships with the node to obtain the positive weight vector corresponding to the node, performing positional addition on the initial weight vectors corresponding to other nodes with negative association relationships with the node to obtain the negative weight vector corresponding to the node, and concatenating the initial weight vector corresponding to the node, the positive weight vector corresponding to the node, and the negative weight vector corresponding to the node to obtain the associated weight vector corresponding to the node, wherein the node is the target account or the associated object; The associated weight vectors of each node are subjected to nonlinear transformation to obtain the weight vectors of the target account and the associated objects.

7. The method according to claim 4, characterized in that, The step of determining the global feature vector of the target account and the global feature vector of the associated objects based on the weight vector corresponding to the target account, the weight vector corresponding to each associated object, the n reference vectors, and the result module includes: The n reference vectors are weighted and summed using the weight vectors corresponding to the target account and the associated object, respectively, to obtain the global feature vector of the target account and the global feature vector of the associated object.

8. The method according to claim 1, characterized in that, The step of determining the local feature vectors of the target account and the associated object based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and a trained local feature extraction model includes: For each node in the node set consisting of the target account and the associated object, the attribute vectors of other nodes with positive association with the node are added side by side to obtain the positive attribute vector of the node, and the feature vectors of other nodes with negative association with the node are added side by side to obtain the negative attribute vector of the node. The attribute vector, the positive attribute vector, and the negative attribute vector of the node are concatenated to obtain the associated attribute vector of the node, wherein the node is the target account or the associated object; The associated attribute vector of each node is subjected to nonlinear transformation to obtain the local feature vector of the target account and the local feature vector of the associated object.

9. The method according to claim 1, characterized in that, The step of determining the feature vectors of the target account and the associated objects based on the global feature vector of the target account, the local feature vector of the target account, the global feature vector of the associated object, the local feature vector of the associated object, and the trained combined model includes: For each node in the node set consisting of the target account and the associated object, the global feature vector and the local feature vector of the node are concatenated to obtain the concatenated feature vector of the node, wherein the node is the target account or the associated object; The concatenated feature vector of each node is subjected to nonlinear transformation to obtain the feature vector of the target account and the feature vector of the associated object.

10. The method according to claim 1, characterized in that, The step of determining the target object to display to the target account based on the currently stored feature vector of the target account and the feature vectors of the multiple objects includes: In the feature vector of the currently stored target account and the feature vectors of the multiple objects, the correlation degree between the feature vector of the target account and the feature vector of each object is determined respectively; Based on the correlation between the feature vector of the target account and the feature vector of each object, the target object to be displayed to the target account is determined.

11. The method according to claim 4, characterized in that, The method further includes: Acquire sample input data and benchmark output data. The sample input data consists of stored attribute vectors of sample accounts, attribute vectors of sample associated objects corresponding to the sample accounts, and the association relationship between sample associated objects and sample accounts. The sample associated objects are objects that have an association relationship with the sample accounts among multiple objects used for display. The benchmark output data includes the association value between each sample associated object and the sample account. The association value of sample associated objects with a positive association relationship with the sample account is 1, and the association value of sample associated objects with a negative association relationship with the sample account is 0. Based on the attribute vector of the sample account, the attribute vector of the sample associated object, the n reference vectors to be trained, and the initialization module to be trained, the initial weight vector corresponding to the sample account and the initial weight vector corresponding to each sample associated object are determined. Based on the initial weight vector corresponding to the sample account, the initial weight vector corresponding to each sample associated object, the association relationship between the sample associated object and the sample account, and the update module to be trained, the weight vector corresponding to the sample account and the weight vector corresponding to the sample associated object are determined. Based on the weight vector corresponding to the sample account, the weight vector corresponding to each sample associated object, the n reference vectors to be trained, and the result module to be trained, the global feature vector of the sample account and the global feature vector of the sample associated object are determined. Based on the attribute vector of the sample account, the attribute vector of the sample associated object, the association relationship between the sample associated object and the sample account, and the local feature extraction model to be trained, the local feature vector of the sample account and the local feature vector of the sample associated object are determined. Based on the global feature vector of the sample account, the local feature vector of the sample account, the global feature vector of the sample associated object, the local feature vector of the sample associated object, and the combined model to be trained, the feature vector of the sample account and the feature vector of the sample associated object are determined. Calculate the correlation degree between the feature vector of each sample account and the feature vector of each sample-associated object to obtain the actual output data; Based on the actual output data and the benchmark output data, the n reference vectors to be trained, the initialization module to be trained, the update module to be trained, the result module to be trained, the local feature extraction model to be trained, and the combined model to be trained are trained to obtain the n reference vectors, the initialization module, the update module, the result module, the local feature extraction model, and the combined model.

12. A device for pushing and displaying information, characterized in that, The device includes: The acquisition module is used to acquire the attribute vector of the target account and the attribute vector of the associated object corresponding to the target account, wherein the associated object is an object that has an association relationship with the target account among multiple objects used for display, and the association relationship is a positive association relationship or a negative association relationship; The determination module determines the feature vector of the target account and the feature vector of the associated object based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and the trained feature extraction model. The feature extraction model includes a global feature extraction model, a local feature extraction model, and a combined model. The step of determining the feature vector of the target account and the feature vector of the associated object based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and the trained feature extraction model includes: Based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and the trained global feature extraction model, the global feature vector of the target account and the global feature vector of the associated object are determined. Based on the attribute vector of the target account, the attribute vector of the associated object, the association relationship between the associated object and the target account, and the trained local feature extraction model, the local feature vector of the target account and the local feature vector of the associated object are determined. Based on the global feature vector of the target account, the local feature vector of the target account, the global feature vector of the associated object, the local feature vector of the associated object, and the trained combined model, the feature vector of the target account and the feature vector of the associated object are determined. The display module is used to determine the target object to be displayed to the target account based on the currently stored feature vector of the target account and the feature vectors of the multiple objects when a display trigger event of the target account is detected, and to push the display information of the target object to the target account.

13. A computer device, characterized in that, The computer device includes a processor and a memory, the memory storing at least one instruction, the at least one instruction being loaded and executed by the processor to perform the operation performed by the method of pushing and displaying information as described in any one of claims 1 to 11.

14. A computer-readable storage medium, characterized in that, The storage medium stores at least one instruction, which is loaded and executed by a processor to perform the operation of the push display information method as described in any one of claims 1 to 11.

Images