Resource push model acquisition, resource push method and storage medium

Abstract

This application discloses a resource push model acquisition method and storage medium, belonging to the field of computer technology. The method includes: processing historical access data of a first object using an initial resource push model to obtain a feature vector of the first object; processing historical access data of multiple second objects using the initial resource push model to obtain multiple feature vectors of the second objects; identifying at least one related object among the multiple second objects based on the feature vectors of the first object and the feature vectors of the multiple second objects; determining a target loss value based on the feature vectors of the first object and the feature vectors of the second objects of at least one related object; and updating the initial resource push model based on the target loss value being greater than a loss value threshold to obtain a target resource push model. The target resource push model obtained by this method can provide more accurate resource pushes to objects, improving the probability of pushed resources being clicked and the effectiveness of resource pushes.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a resource push model acquisition, resource push method, and storage medium. Background Technology

[0002] With the rapid development of computer technology, resource recommendation platforms have also developed rapidly. These platforms infer user preferences based on historical information, determine target resources to recommend to the user based on those preferences, and then push those target resources to the user.

[0003] However, when a user has limited historical information, the matching degree between the specific push resources and the user is low, resulting in a lower probability that the push resources will be clicked by the user. Consequently, the accuracy, reliability, and effectiveness of resource push are all lower. Summary of the Invention

[0004] This application provides an embodiment of a resource push model acquisition, a resource push method, and a storage medium, which can be used to solve problems in related technologies. The technical solution is as follows:

[0005] In a first aspect, embodiments of this application provide a method for obtaining a resource push model, the method comprising:

[0006] The initial resource push model is invoked to process the historical access data of the first object to obtain the feature vector of the first object, which is used to characterize the first object.

[0007] The initial resource push model is invoked to process the historical access data of multiple second objects, resulting in multiple feature vectors of the second objects. Any feature vector of a second object is used to characterize any second object.

[0008] Based on the first object feature vector and the plurality of second object feature vectors, at least one related object is determined among the plurality of second objects, wherein the related object is an object whose correlation with the first object meets the correlation requirements;

[0009] A target loss value is determined based on the feature vector of the first object and the feature vector of the second object of the at least one related object, the target loss value being used to indicate the accuracy of the initial resource push model;

[0010] Based on the fact that the target loss value is greater than the loss value threshold, the initial resource push model is updated to obtain the target resource push model.

[0011] In one possible implementation, the step of calling the initial resource push model to process the historical access data of the first object to obtain the feature vector of the first object includes:

[0012] Based on the fact that the historical access data of the first object is one, the initial resource push model is called to process the historical access data of the first object to obtain the feature vector of the historical access data of the first object, and the feature vector of the historical access data of the first object is used as the feature vector of the first object.

[0013] Since there are multiple historical access data for the first object, target historical access data is determined from the multiple historical access data of the first object. The initial resource push model is called to process the target historical access data to obtain the feature vector of the target historical access data. The feature vector of the target historical access data is used as the feature vector of the first object.

[0014] In one possible implementation, the step of invoking the initial resource push model to process the historical access data of the first object to obtain a feature vector of the historical access data of the first object includes:

[0015] Based on the historical access data of the first object, including the resource text, resource identifier and resource image of the historical access resources of the first object, the initial resource push model includes an initial text vector determination sub-model, an initial identifier vector determination sub-model and an initial image vector determination sub-model. The initial text vector determination sub-model is called to process the resource text of the historical access resources of the first object to obtain the text feature vector of the historical access resources of the first object.

[0016] The initial image vector determination sub-model is called to process the resource images of the historical accessed resources of the first object to obtain the image feature vector of the historical accessed resources of the first object;

[0017] The initial identifier vector determination sub-model is invoked to process the resource identifier of the historical accessed resources of the first object, the text feature vector, and the image feature vector to obtain the identifier feature vector of the historical accessed resources of the first object;

[0018] Based on the text feature vector, image feature vector, and identifier feature vector of the historical access resources of the first object, the feature vector of the historical access data of the first object is determined.

[0019] In one possible implementation, determining the target historical access data from multiple historical access data of the first object includes:

[0020] Obtain the access times of multiple historical access data for the first object;

[0021] The historical access data that meets the time requirement among the multiple historical access data of the first object is taken as the target historical access data.

[0022] In one possible implementation, determining at least one related object among the plurality of second objects based on the first object feature vector and the plurality of second object feature vectors includes:

[0023] Based on the feature vector of the first object and the feature vectors of the plurality of second objects, the similarity between each second object and the first object is determined.

[0024] The second object whose similarity meets the similarity requirement is considered as the related object.

[0025] In one possible implementation, determining the target loss value based on the first object feature vector and the second object feature vector of the at least one related object includes:

[0026] Based on the feature vector of the first object and the feature vector of the second object of the at least one related object, the weight of each related object is determined, and the weight of any related object is used to indicate the influence of any related object on the first object;

[0027] Based on the second object feature vectors of each related object and the weights of each related object, a first target feature vector is determined, which is used to characterize the interest preferences of the first object.

[0028] Based on the historical access data of the first object and the historical access data of the third object, a second target feature vector is determined. The second target feature vector is used to characterize the interest preferences of the first object. The third object is any one of the at least one related object.

[0029] The target loss value is determined based on the first target feature vector and the second target feature vector.

[0030] In one possible implementation, the method further includes:

[0031] A second loss value is determined based on the feature vector of the first object, the resource feature vector of the first resource, and the resource feature vector of the second resource. The first resource is the resource that the first object has historically accessed, and the second resource is the resource that the first object has not accessed.

[0032] Determining the target loss value based on the first target feature vector and the second target feature vector includes:

[0033] Based on the first target feature vector and the second target feature vector, determine the first loss value;

[0034] The target loss value is determined based on the first loss value and the second loss value.

[0035] In one possible implementation, the initial resource push model includes an initial transformer sub-model, which includes N layers, where N is a positive integer;

[0036] The step of determining the second target feature vector based on the historical access data of the first object and the historical access data of the third object includes:

[0037] Based on the historical access data of the first object, including the resource text, resource identifier, and resource image of the historical access resources of the first object, the feature vector of the historical access data of the first object in the M-th layer of the initial transformer sub-model is determined according to the resource text, resource identifier, and resource image of the historical access resources of the first object, where M is greater than 0 and less than N;

[0038] Based on the historical access data of the third object, including the resource text, resource identifier, and resource image of the historical access resources of the third object, the feature vector of the historical access data of the third object in the M-th layer of the initial transformer sub-model is determined according to the resource text, resource identifier, and resource image of the historical access resources of the third object.

[0039] Based on the feature vector of the historical access data of the first object in the Mth layer of the initial converter sub-model and the feature vector of the historical access data of the third object in the Mth layer of the initial converter sub-model, a fused feature vector is determined;

[0040] The fused feature vector is updated through layers M+1 to N of the initial transformer sub-model to obtain the second target feature vector.

[0041] In one possible implementation, determining the feature vector of the historical access data of the first object at the M-th layer of the initial transformer sub-model based on the resource text, resource identifier, and resource image of the resources accessed by the first object in its history includes:

[0042] Based on the resource text, resource identifier, and resource image of the first object's historical accessed resources, determine the reference text feature vector of the resource text of the first object's historical accessed resources in the M-th layer of the initial converter sub-model, the reference identifier feature vector of the resource identifier of the first object's historical accessed resources in the M-th layer of the initial converter sub-model, and the reference image feature vector of the resource image of the first object's historical accessed resources in the M-th layer of the initial converter sub-model.

[0043] The reference text feature vector, the reference identifier feature vector, and the reference image feature vector are fused to obtain the feature vector of the historical access data of the first object at the Mth layer of the initial transformer sub-model.

[0044] Secondly, embodiments of this application provide a resource push method, the method comprising:

[0045] Obtain historical access data of the target object and a target resource push model, wherein the target resource push model is obtained based on any of the methods described in the first aspect above;

[0046] The target resource push model is invoked to process the historical access data of the target object, resulting in multiple target resources;

[0047] The multiple target resources are pushed to the target object.

[0048] In one possible implementation, the process of calling the target resource push model to process the historical access data of the target object yields multiple target resources, including:

[0049] The target resource push model is invoked to process the historical access data of the target object to obtain the object feature vector of the target object, which is used to characterize the target object.

[0050] Based on the object feature vector of the target object and the resource feature vectors of multiple candidate resources, multiple target resources are determined from the multiple candidate resources.

[0051] Thirdly, embodiments of this application provide a resource push model acquisition device, the device comprising:

[0052] The processing module is used to call the initial resource push model to process the historical access data of the first object and obtain the feature vector of the first object. The feature vector of the first object is used to characterize the first object.

[0053] The processing module is also used to call the initial resource push model to process the historical access data of multiple second objects to obtain multiple feature vectors of second objects, and any feature vector of a second object is used to characterize any second object.

[0054] The determining module is configured to determine at least one related object among the plurality of second objects based on the feature vector of the first object and the feature vectors of the plurality of second objects, wherein the related object is an object whose correlation with the first object meets the correlation requirements;

[0055] The determining module is further configured to determine a target loss value based on the first object feature vector and the second object feature vector of the at least one related object, wherein the target loss value is used to indicate the accuracy of the initial resource push model;

[0056] The update module is used to update the initial resource push model based on the target loss value being greater than the loss value threshold, so as to obtain the target resource push model.

[0057] In one possible implementation, the processing module is configured to: 1) Based on the fact that the historical access data of the first object is singular, call the initial resource push model to process the historical access data of the first object, obtain a feature vector of the historical access data of the first object, and use the feature vector of the historical access data of the first object as the feature vector of the first object; 2) Based on the fact that the historical access data of the first object is multiple, determine a target historical access data among the multiple historical access data of the first object, call the initial resource push model to process the target historical access data, obtain a feature vector of the target historical access data, and use the feature vector of the target historical access data as the feature vector of the first object.

[0058] In one possible implementation, the processing module is configured to, based on the historical access data of the first object, including resource text, resource identifier, and resource image of the historical accessed resources of the first object, and the initial resource push model including an initial text vector determination sub-model, an initial identifier vector determination sub-model, and an initial image vector determination sub-model, call the initial text vector determination sub-model to process the resource text of the historical accessed resources of the first object to obtain a text feature vector of the historical accessed resources of the first object; call the initial image vector determination sub-model to process the resource image of the historical accessed resources of the first object to obtain an image feature vector of the historical accessed resources of the first object; call the initial identifier vector determination sub-model to process the resource identifier, the text feature vector, and the image feature vector of the historical accessed resources of the first object to obtain an identifier feature vector of the historical accessed resources of the first object; and determine the feature vector of the historical access data of the first object based on the text feature vector, image feature vector, and identifier feature vector of the historical accessed resources of the first object.

[0059] In one possible implementation, the processing module is configured to obtain the access times of multiple historical access data of the first object; and select the historical access data whose access time meets the time requirement from the multiple historical access data of the first object as the target historical access data.

[0060] In one possible implementation, the determining module is configured to determine the similarity between each second object and the first object based on the feature vector of the first object and the feature vectors of the plurality of second objects; and to designate the second objects whose similarity meets the similarity requirements as the related objects.

[0061] In one possible implementation, the determining module is configured to: determine the weights of each related object based on the first object feature vector and the second object feature vectors of the at least one related object, wherein the weight of any related object indicates the influence of any related object on the first object; determine a first target feature vector based on the second object feature vectors of each related object and the weights of each related object, wherein the first target feature vector characterizes the interest preference of the first object; determine a second target feature vector based on the historical access data of the first object and the historical access data of a third object, wherein the second target feature vector characterizes the interest preference of the first object, and the third object is any one of the at least one related object; and determine the target loss value based on the first target feature vector and the second target feature vector.

[0062] In one possible implementation, the determining module is further configured to determine a second loss value based on the feature vector of the first object, the resource feature vector of the first resource, and the resource feature vector of the second resource, wherein the first resource is a resource that the first object has historically accessed, and the second resource is a resource that the first object has not accessed.

[0063] The determining module is configured to determine a first loss value based on the first target feature vector and the second target feature vector; and to determine the target loss value based on the first loss value and the second loss value.

[0064] In one possible implementation, the initial resource push model includes an initial transformer sub-model, which includes N layers, where N is a positive integer;

[0065] The determining module is configured to: determine the feature vector of the historical access data of the first object at the M-th layer of the initial converter sub-model based on the historical access data of the first object, including the resource text, resource identifier, and resource image of the historical accessed resources of the first object, where M is greater than 0 and less than N; determine the feature vector of the historical access data of the third object at the M-th layer of the initial converter sub-model based on the historical access data of the third object, including the resource text, resource identifier, and resource image of the historical accessed resources of the third object; determine a fused feature vector based on the feature vector of the historical access data of the first object at the M-th layer of the initial converter sub-model and the feature vector of the historical access data of the third object at the M-th layer of the initial converter sub-model; and update the fused feature vector through the (M+1)-th to (N)-th layers of the initial converter sub-model to obtain the second target feature vector.

[0066] In one possible implementation, the determining module is configured to determine, based on the resource text, resource identifier, and resource image of the historically accessed resources of the first object, a reference text feature vector of the resource text of the historically accessed resources of the first object at the M-th layer of the initial converter sub-model, a reference identifier feature vector of the resource identifier of the historically accessed resources of the first object at the M-th layer of the initial converter sub-model, and a reference image feature vector of the resource image of the historically accessed resources of the first object at the M-th layer of the initial converter sub-model; and to fuse the reference text feature vector, the reference identifier feature vector, and the reference image feature vector to obtain the feature vector of the historical access data of the first object at the M-th layer of the initial converter sub-model.

[0067] Fourthly, embodiments of this application provide a resource push device, the device comprising:

[0068] The acquisition module is used to acquire historical access data of the target object and the target resource push model, wherein the target resource push model is obtained based on any of the devices described in the third aspect above.

[0069] The processing module is used to call the target resource push model to process the historical access data of the target object to obtain multiple target resources;

[0070] The push module is used to push the multiple target resources to the target object.

[0071] In one possible implementation, the processing module is used to call the target resource push model to process the historical access data of the target object to obtain the object feature vector of the target object, the object feature vector of the target object is used to characterize the target object; and based on the object feature vector of the target object and the resource feature vectors of multiple candidate resources, multiple target resources are determined from the multiple candidate resources.

[0072] Fifthly, embodiments of this application provide an electronic device, which includes a processor and a memory. The memory stores at least one piece of program code, which is loaded and executed by the processor to enable the electronic device to implement the resource push model acquisition method described in the first aspect or any possible implementation of the first aspect, or to enable the electronic device to implement the resource push method described in the second aspect or any possible implementation of the second aspect.

[0073] In a sixth aspect, a computer-readable storage medium is also provided, wherein at least one piece of program code is stored therein, the at least one piece of program code being loaded and executed by a processor to enable a computer to implement the resource push model acquisition method described in the first aspect or any possible implementation thereof, or to enable an electronic device to implement the resource push method described in the second aspect or any possible implementation thereof.

[0074] In a seventh aspect, a computer program or computer program product is also provided, wherein the computer program or computer program product stores at least one computer instruction, the at least one computer instruction being loaded and executed by a processor to enable a computer to implement the resource push model acquisition method described in the first aspect or any possible implementation of the first aspect, or to enable an electronic device to implement the resource push method described in the second aspect or any possible implementation of the second aspect.

[0075] The technical solution provided in this application has at least the following beneficial effects:

[0076] The technical solution provided in this application update the initial resource push model by considering not only the historical access data of the first object but also the historical access data of the second object. By identifying related objects in the second object that are related to the first object, the initial resource push model is updated to ensure that the accuracy of the updated target resource push model is high. This enables more accurate resource push to the object, resulting in a higher matching degree between the pushed resource and the object, increasing the probability that the pushed resource is clicked by the object, and thus improving the effect of resource push. Attached Figure Description

[0077] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0078] Figure 1 This is a schematic diagram of the implementation environment for obtaining a resource push model and a resource push method provided in an embodiment of this application;

[0079] Figure 2 This is a flowchart of a method for obtaining a resource push model provided in an embodiment of this application;

[0080] Figure 3 This is a schematic diagram illustrating the process of obtaining the feature vector of historical access data of a first object according to an embodiment of this application;

[0081] Figure 4 This is a schematic diagram illustrating the process of obtaining an initial text vector determination sub-model according to an embodiment of this application;

[0082] Figure 5 This is a schematic diagram illustrating the process of obtaining an initial image vector to determine a sub-model, as provided in an embodiment of this application.

[0083] Figure 6 This is a schematic diagram illustrating the process of obtaining an initial identifier vector to determine a sub-model, as provided in an embodiment of this application.

[0084] Figure 7 This is a flowchart illustrating the determination of a second target feature vector according to an embodiment of this application;

[0085] Figure 8 This is a flowchart of a resource push method provided in an embodiment of this application;

[0086] Figure 9 This is a schematic diagram of the structure of a resource push model acquisition device provided in an embodiment of this application;

[0087] Figure 10 This is a schematic diagram of the structure of a resource push device provided in an embodiment of this application;

[0088] Figure 11 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application;

[0089] Figure 12 This application provides a schematic diagram of the structure of a server. Detailed Implementation

[0090] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0091] It should be noted that the terms "first," "second," etc., used in this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0092] Figure 1 This is a schematic diagram illustrating the implementation environment of a resource push model acquisition and resource push method provided in an embodiment of this application, such as... Figure 1 As shown, the implementation environment includes: terminal device 101 and server 102.

[0093] The resource push model acquisition method provided in this application embodiment can be executed by terminal device 101, server 102, or jointly by terminal device 101 and server 102. This application embodiment does not limit this. In the case where the resource push model acquisition method provided in this application embodiment is jointly executed by terminal device 101 and server 102, server 102 undertakes the main computational work, and terminal device 101 undertakes the secondary computational work; or, server 102 undertakes the secondary computational work, and terminal device 101 undertakes the main computational work; or, server 102 and terminal device 101 use a distributed computing architecture for collaborative computation.

[0094] The resource push method provided in this application embodiment can be executed by terminal device 101, server 102, or jointly by terminal device 101 and server 102. This application embodiment does not limit this. In the case where the resource push method provided in this application embodiment is jointly executed by terminal device 101 and server 102, server 102 undertakes the main computational work, and terminal device 101 undertakes the secondary computational work; or, server 102 undertakes the secondary computational work, and terminal device 101 undertakes the main computational work; or, server 102 and terminal device 101 use a distributed computing architecture for collaborative computation.

[0095] It should be noted that the execution device of the resource push model acquisition method and the execution device of the resource push method can be the same or different, and this application embodiment does not limit this. For example, the execution device of the resource push model acquisition method is terminal device 101, and the execution device of the resource push method is server 102; or, both the execution device of the resource push model acquisition method and the execution device of the resource push method are terminal device 101.

[0096] Optionally, the terminal device 101 can be any electronic product capable of human-computer interaction with a user through one or more methods such as a keyboard, touchpad, touchscreen, remote control, voice interaction, or handwriting device. The terminal device 101 includes, but is not limited to, mobile phones, computers, intelligent voice interaction devices, smart home appliances, vehicle terminals, and aircraft. The server 102 can be a single server, a server cluster consisting of multiple servers, or any of the following: a cloud computing platform or a virtualization center. This application embodiment does not limit this. The server 102 communicates with the terminal device 101 via a wired or wireless network. The server 102 has data receiving, data processing, and data sending functions. Of course, the server 102 may also have other functions, which are not limited in this application embodiment.

[0097] Those skilled in the art should understand that the terminal device 101 and server 102 described above are merely illustrative examples. Other existing or future terminal devices or servers that are applicable to this application should also be included within the scope of protection of this application, and are hereby incorporated by reference.

[0098] This application provides a method for obtaining a resource push model, which can be applied to the above-described implementation environment. Figure 2 The flowchart shown in this embodiment of the present application illustrates a method for obtaining a resource push model. This method can be implemented by... Figure 1 The electronic device in the process can be executed by the electronic device. Figure 1 Terminal device 101 in the middle can also be Figure 1 The server 102 in this embodiment is not limited thereto. Figure 2 As shown, the method includes the following steps 201 to 205.

[0099] In step 201, the initial resource push model is invoked to process the historical access data of the first object to obtain the feature vector of the first object, which is used to characterize the first object.

[0100] In the exemplary embodiments of this application, historical access data includes, but is not limited to, at least one of the following: resource text, resource identifier, and resource image of historically accessed resources. Resource text describes the resource; the resource identifier can be a resource number, a resource name, or any identifier that represents the resource (this embodiment does not limit this); and the resource image is an image of the resource. The historical access resources of the first object include, but are not limited to, at least one of the following: resources historically browsed by the first object, resources historically clicked by the first object, resources historically converted by the first object, and resources historically searched by the first object. For example, the historical access data of the first object includes: product A, description text of product A, and image of product A.

[0101] In one possible implementation, the electronic device acquires a training dataset and an initial resource push model. The training dataset includes historical access data for multiple objects. Any one of the multiple objects is selected as the first object, and its historical access data is then acquired. The initial resource push model includes at least one of an initial text vector determination sub-model, an initial identifier vector determination sub-model, and an initial image vector determination sub-model, and an initial transformer sub-model. The initial text vector determination sub-model includes an embedding layer, a self-attention layer, and a prediction layer. The initial identifier vector determination sub-model includes an embedding layer, a feature-aware self-attention layer, and a prediction layer. The initial image vector determination sub-model includes an embedding layer, a self-attention layer, and a prediction layer.

[0102] The process of processing the historical access data of the first object using the initial resource push model to obtain the feature vector of the first object includes: If the historical access data of the first object is singular, the initial resource push model is called to process the historical access data of the first object to obtain the feature vector of the historical access data of the first object; the feature vector of the historical access data of the first object is used as the feature vector of the first object. Alternatively, if the historical access data of the first object is multiple, a target historical access data is determined from the multiple historical access data of the first object; the initial resource push model is called to process the target historical access data to obtain the feature vector of the target historical access data; the feature vector of the target historical access data is used as the feature vector of the first object.

[0103] The process of calling the initial resource push model to process the historical access data of the first object and obtaining the feature vector of the historical access data of the first object is similar to the process of calling the initial resource push model to process the target historical access data and obtaining the feature vector of the target historical access data. This application embodiment only uses the process of calling the initial resource push model to process the historical access data of the first object and obtaining the feature vector of the historical access data of the first object as an example for explanation.

[0104] In one possible implementation, the process of obtaining the feature vector of the historical access data of the first object will differ depending on the content included in the historical access data of the first object and the sub-models included in the initial resource push model. This application embodiment uses only the following seven implementation methods to obtain the feature vector of the historical access data of the first object.

[0105] Implementation Method 1: Based on the historical access data of the first object, including the resource text of the historical accessed resources of the first object, the initial resource push model includes an initial text vector determination sub-model. The initial text vector determination sub-model is called to process the resource text of the historical accessed resources of the first object to obtain the text feature vector of the historical accessed resources of the first object; the text feature vector of the historical accessed resources of the first object is used as the feature vector of the historical access data of the first object.

[0106] Implementation Method 2: Based on the historical access data of the first object, including the resource identifier of the historical accessed resources of the first object, the initial resource push model includes an initial identifier vector determination sub-model. The initial identifier vector determination sub-model is called to process the resource identifier of the historical accessed resources of the first object to obtain the identifier feature vector of the historical accessed resources of the first object; the identifier feature vector of the historical accessed resources of the first object is used as the feature vector of the historical access data of the first object.

[0107] Implementation Method 3: Based on the historical access data of the first object, including the resource images of the historical accessed resources of the first object, the initial resource push model includes the initial image vector determination sub-model. The initial image vector determination sub-model is called to process the resource images of the historical accessed resources of the first object to obtain the image feature vector of the historical accessed resources of the first object; the image feature vector of the historical accessed resources of the first object is used as the feature vector of the historical access data of the first object.

[0108] Implementation Method 4: Based on the historical access data of the first object, including the resource text and resource identifier of the historical accessed resources of the first object, the initial resource push model includes an initial text vector determination sub-model and an initial identifier vector determination sub-model. The initial text vector determination sub-model is called to process the resource text of the historical accessed resources of the first object to obtain the text feature vector of the historical accessed resources of the first object; the initial identifier vector determination sub-model is called to process the resource identifier of the historical accessed resources of the first object to obtain the identifier feature vector of the historical accessed resources of the first object; based on the text feature vector and the identifier feature vector of the historical accessed resources of the first object, the feature vector of the historical access data of the first object is determined. Optionally, the average value of the text feature vector and the identifier feature vector of the historical accessed resources of the first object is used as the feature vector of the historical access data of the first object.

[0109] Implementation Method 5: Based on the historical access data of the first object, including the resource text and resource images of the historical accessed resources of the first object, the initial resource push model includes an initial text vector determination sub-model and an initial image vector determination sub-model. The initial text vector determination sub-model is called to process the resource text of the historical accessed resources of the first object to obtain the text feature vector of the historical accessed resources of the first object; the initial image vector determination sub-model is called to process the resource images of the historical accessed resources of the first object to obtain the image feature vector of the historical accessed resources of the first object; based on the text feature vector and image feature vector of the historical accessed resources of the first object, the feature vector of the historical access data of the first object is determined. Optionally, the average value of the text feature vector and image feature vector of the historical accessed resources of the first object is used as the feature vector of the historical access data of the first object.

[0110] Implementation Method Six: Based on the historical access data of the first object, including the resource identifier and resource image of the historical accessed resources of the first object, the initial resource push model includes an initial identifier vector determination sub-model and an initial image vector determination sub-model. The initial image vector determination sub-model is called to process the resource images of the historical accessed resources of the first object, obtaining the image feature vector of the historical accessed resources of the first object. The initial identifier vector determination sub-model is called to process the resource identifier of the historical accessed resources of the first object, obtaining the identifier feature vector of the historical accessed resources of the first object. Based on the image feature vector and the identifier feature vector of the historical accessed resources of the first object, the feature vector of the historical access data of the first object is determined. Optionally, the average value of the image feature vector and the identifier feature vector of the historical accessed resources of the first object is used as the feature vector of the historical access data of the first object.

[0111] Implementation Method Seven: Based on the historical access data of the first object, including the resource text, resource identifier, and resource image of the historical accessed resources of the first object, the initial resource push model includes an initial text vector determination sub-model, an initial identifier vector determination sub-model, and an initial image vector determination sub-model. The initial text vector determination sub-model is called to process the resource text of the historical accessed resources of the first object, obtaining the text feature vector of the historical accessed resources of the first object; the initial image vector determination sub-model is called to process the resource image of the historical accessed resources of the first object, obtaining the image feature vector of the historical accessed resources of the first object; the initial identifier vector determination sub-model is called to process the resource identifier, text feature vector, and image feature vector of the historical accessed resources of the first object, obtaining the identifier feature vector of the historical accessed resources of the first object; based on the text feature vector, image feature vector, and identifier feature vector of the historical accessed resources of the first object, the feature vector of the historical access data of the first object is determined. For example... Figure 3 This is a schematic diagram illustrating the process of obtaining the feature vector of historical access data of a first object according to an embodiment of this application. This process is consistent with that of implementation method seven and will not be described again here.

[0112] The process of calling the initial identifier vector determination sub-model to process the resource identifier, text feature vector, and image feature vector of the historical accessed resources of the first object to obtain the identifier feature vector of the historical accessed resources of the first object includes: calling the initial identifier vector determination sub-model to process the resource identifier of the historical accessed resources of the first object to obtain a reference feature vector, and using the average value of the reference feature vector, text feature vector, and image feature vector as the identifier feature vector of the historical accessed resources of the first object.

[0113] Optionally, the initial identifier vector can be called to determine the resource identifier of the historical accessed resources of the first object to process the resource identifier, obtain the reference feature vector, and use the reference feature vector as the identifier feature vector of the historical accessed resources of the first object.

[0114] The process of determining the feature vector of the historical access data of the first object based on the text feature vector, image feature vector and identifier feature vector of the historical access resources of the first object includes: taking the average value of the text feature vector, image feature vector and identifier feature vector of the historical access resources of the first object as the feature vector of the historical access data of the first object.

[0115] The feature vector of the historical access data of the first object can be obtained by any of the above implementation methods, and the embodiments of this application do not limit this.

[0116] For example, if the text feature vector of the historical accessed resource of the first object is (A1, A2, A3, A4), the image feature vector of the historical accessed resource of the first object is (B1, B2, B3, B4), and the identifier feature vector of the historical accessed resource of the first object is (C1, C2, C3, C4), then the feature vector of the historical access data of the first object is...

[0117] In one possible implementation, before calling the initial text vector determination sub-model to process the resource text of the first object's historical access resources and obtain the text feature vector of the first object's historical access resources, it is necessary to first obtain the initial text vector determination sub-model, such as... Figure 4 This is a schematic diagram illustrating the process of obtaining an initial text vector determination sub-model according to an embodiment of this application. The process includes: obtaining training text, standard text feature vectors of the training text, and a first text vector determination sub-model; processing the training text using the first text vector determination sub-model to obtain predicted text feature vectors of the training text; determining a text loss value based on the standard text feature vectors and predicted text feature vectors of the training text; updating the first text vector determination sub-model based on a text loss value greater than a text loss threshold to obtain the initial text vector determination sub-model; and using the first text vector determination sub-model as the initial text vector determination sub-model based on a text loss value not greater than a text loss threshold. The text loss threshold is set based on experience or adjusted according to the implementation environment; this embodiment of the application does not limit this setting.

[0118] Before calling the initial image vector determination sub-model to process the resource images of the first object's historical access resources and obtain the image feature vectors of the first object's historical access resources, it is necessary to first obtain the initial image vector determination sub-model. The process of obtaining the initial image vector determination sub-model is similar to the process of obtaining the initial text vector determination sub-model described above. Figure 5This is a schematic diagram illustrating the process of obtaining an initial image vector determination sub-model according to an embodiment of this application. The process includes: obtaining training images, standard image feature vectors of the training images, and a first image vector determination model; processing the training images using the first image vector determination sub-model to obtain predicted image feature vectors for the training images; determining an image loss value based on the standard image feature vectors and the predicted image feature vectors of the training images; updating the first image vector determination sub-model based on the image loss value being greater than an image loss threshold to obtain the initial image vector determination sub-model; and using the first image vector determination sub-model as the initial image vector determination sub-model based on the image loss value not being greater than the image loss threshold. The image loss threshold is set based on experience or adjusted according to the implementation environment; this embodiment of the application does not limit this.

[0119] Before calling the initial identifier vector determination sub-model to process the resource identifiers of the first object's historically accessed resources and obtain the identifier feature vectors of the first object's historically accessed resources, it is necessary to first obtain the initial identifier vector determination sub-model. The process of obtaining the initial identifier vector determination sub-model is similar to the process of obtaining the initial text vector determination sub-model described above. For example... Figure 6 This is a schematic diagram illustrating the process of obtaining an initial identifier vector determination sub-model according to an embodiment of this application. The process includes: obtaining training identifiers, standard identifier feature vectors of the training identifiers, and a first identifier vector determination model; processing the training identifiers using the first identifier vector determination sub-model to obtain predicted identifier feature vectors of the training identifiers; determining identifier loss values ​​based on the standard identifier feature vectors and predicted identifier feature vectors of the training identifiers; updating the first identifier vector determination sub-model based on the identifier loss value being greater than an identifier loss threshold to obtain the initial identifier vector determination sub-model; and using the first identifier vector determination sub-model as the initial identifier vector determination sub-model based on the identifier loss value not being greater than the identifier loss threshold. The identifier loss threshold is set based on experience or adjusted according to the implementation environment; this embodiment of the application does not limit this.

[0120] In one possible implementation, the process of determining the target historical access data from multiple historical access data of the first object includes: obtaining the access times of multiple historical access data of the first object, and selecting the historical access data whose access times meet the required interval as the target historical access data. For example, since the most recent historical access data of the first object can indicate the interests and preferences of the first object, the historical access data with the smallest time difference from the current time among the multiple historical access data of the first object can be selected as the target historical access data.

[0121] In step 202, the initial resource push model is invoked to process the historical access data of multiple second objects to obtain multiple feature vectors of second objects. Any feature vector of a second object is used to characterize any second object.

[0122] In one possible implementation, objects in the training dataset other than the first object are used as second objects, thus obtaining historical access data for multiple second objects. The process of calling the initial resource push model to process the historical access data of multiple second objects to obtain feature vectors for multiple second objects is similar to the process of calling the initial resource push model to process the historical access data of the first object to obtain feature vectors for the first object in step 201 above, and will not be described again here.

[0123] In step 203, at least one related object is determined among the multiple second objects based on the feature vector of the first object and the feature vectors of multiple second objects.

[0124] Among them, the relevant objects are those whose relevance to the first object meets the relevance requirements, that is, objects with similar interests and preferences to the first object.

[0125] In one possible implementation, the process of determining at least one related object among multiple second objects based on the feature vector of a first object and multiple feature vectors of second objects includes: determining the similarity between each second object and the first object based on the feature vector of the first object and the feature vectors of multiple second objects; and identifying the second objects whose similarity meets the similarity requirement as related objects. Here, "meeting the similarity requirement" means that the similarity is greater than a similarity threshold. The similarity threshold is set based on experience or adjusted according to the implementation environment, and this application embodiment does not limit this. For example, the similarity threshold is 90. Optionally, "meeting the similarity requirement" can also mean that after sorting the similarities between each second object and the first object in descending order, the second object with the highest similarity before the target position is identified. Here, the target position is set based on experience or adjusted according to the implementation environment, and this application embodiment does not limit this. For example, the target position is 6.

[0126] Optionally, the process of determining the similarity between each second object and the first object based on the feature vector of the first object and the feature vectors of multiple second objects includes: for any one of the multiple second objects, taking the dot product of the feature vector of the first object and the feature vector of the second object of any one of the second objects as the similarity between any one of the second objects and the first object.

[0127] For example, based on the feature vector of the first object and the feature vectors of multiple second objects, the similarity between each second object and the first object is determined according to the following formula (1).

[0128]

[0129] In the above formula (1), r(u, z) j Let v be the similarity between the first object and the j-th second object. u The first object's feature vector. Let be the feature vector of the j-th second object.

[0130] For example, if the feature vector of the first object is (1, 2, 3, 4), and the feature vector of any second object is (2, 1, 4, 3), then the similarity between any second object and the first object is 28.

[0131] In step 204, a target loss value is determined based on the first object feature vector and the second object feature vector of at least one related object. The target loss value is used to indicate the accuracy of the initial resource push model.

[0132] Optionally, the process of determining the target loss value based on the feature vector of the first object and the feature vector of the second object of at least one related object includes: determining the weight of each related object based on the feature vector of the first object and the feature vector of the second object of at least one related object, wherein the weight of any related object is used to indicate the influence of any related object on the first object; determining the first target feature vector based on the feature vector of the second object of each related object and the weight of each related object; determining the second target feature vector based on the historical access data of the first object and the historical access data of the third object; and determining the target loss value based on the first target feature vector and the second target feature vector. Wherein, the first target feature vector is used to characterize the interest preference of the first object, the second target feature vector is used to characterize the interest preference of the first object, and the third object is any one of the at least one related object.

[0133] In one possible implementation, the process of determining the weights of each related object based on the first object feature vector and the second object feature vector of at least one related object includes: for any one of the at least one related objects, determining a first vector based on the first object feature vector, determining a second vector based on the second object feature vector of any one related object, and determining the weight of any one related object based on the first vector, the second vector, and a target value, wherein the target value is the dimension of the first object feature vector, and the dimension of the first object feature vector is the same as the dimension of the second object feature vector.

[0134] Optionally, the weights of each related object are determined according to the following formula (2) based on the feature vector of the first object and the feature vector of the second object of at least one related object.

[0135]

[0136] In formula (2) above, α is the weight of any relevant object, softmax is the activation function, and v u Z is the feature vector of the first object. u Let be the second object feature vector of any related object, d be the dimension of the first object feature vector, and T be the transpose operation. (v u ·W1) T The first value, (Z) u •W2) is the second value. W1 and W2 are weight parameters, which are set based on experience or adjusted according to the implementation environment. This application does not limit this.

[0137] In one possible implementation, after determining the weights of each relevant object, the first target feature vector is determined according to the second object feature vector of each relevant object and the weights of each relevant object, in accordance with the following formula (3).

[0138]

[0139] In the above formula (3), Z is the first target feature vector, Z includes the second object feature vectors of each related object, and ω includes the weights of each related object.

[0140] In one possible implementation, the initial resource recommendation model includes an initial transformer sub-model, which consists of N layers, where N is a positive integer. The process of determining the second target feature vector based on the historical access data of the first object and the third object includes: Based on the historical access data of the first object, including the resource text, resource identifier, and resource image of the resources accessed historically by the first object, determine the feature vector of the historical access data of the first object at layer M of the initial converter sub-model, where M is greater than 0 and less than N; Based on the historical access data of the third object, including the resource text, resource identifier, and resource image of the resources accessed historically by the third object, determine the feature vector of the historical access data of the third object at layer M of the initial converter sub-model; Based on the feature vectors of the historical access data of the first object and the third object at layer M of the initial converter sub-model, determine the fused feature vector; Update the fused feature vector through layers M+1 to N of the initial converter sub-model to obtain the second target feature vector.

[0141] The process of determining the fused feature vector based on the feature vectors of the historical access data of the first object in the Mth layer of the initial converter sub-model and the feature vectors of the historical access data of the third object in the Mth layer of the initial converter sub-model includes: taking the average of the feature vectors of the historical access data of the first object in the Mth layer of the initial converter sub-model and the feature vectors of the historical access data of the third object in the Mth layer of the initial converter sub-model as the fused feature vector.

[0142] The process described above for determining the feature vector of the historical access data of the first object in the Mth layer of the initial converter sub-model based on the resource text, resource identifier, and resource image of the historical accessed resources of the first object is similar to the process described above for determining the feature vector of the historical access data of the third object in the Mth layer of the initial converter sub-model based on the resource text, resource identifier, and resource image of the historical accessed resources of the third object. This embodiment of the application will only use the process of determining the feature vector of the historical access data of the first object in the Mth layer of the initial converter sub-model based on the resource text, resource identifier, and resource image of the historical accessed resources of the first object as an example for explanation.

[0143] The process of determining the feature vector of the historical access data of the first object at the Mth layer of the initial converter sub-model based on the resource text, resource identifier, and resource image of the historical accessed resources of the first object includes: determining the reference text feature vector of the resource text of the historical accessed resources at the Mth layer of the initial converter sub-model, the reference identifier feature vector of the resource identifier of the historical accessed resources at the Mth layer of the initial converter sub-model, and the reference image feature vector of the resource image of the historical accessed resources at the Mth layer of the initial converter sub-model; fusing the reference text feature vector, the reference identifier feature vector, and the reference image feature vector to obtain the feature vector of the historical access data of the first object at the Mth layer of the initial converter sub-model. Optionally, the average value of the reference text feature vector, the reference identifier feature vector, and the reference image feature vector is used as the feature vector of the historical access data of the first object at the Mth layer of the initial converter sub-model.

[0144] Formula (4) below is the process for determining the text feature vector of the resource text of the first object's historical access resource in the Mth layer of the initial transformer sub-model, as provided in the embodiments of this application.

[0145]

[0146] In the above formula (4), Let gl be the text feature vector of the resource text of the first object's historical access to resources at the M-th layer of the initial transformer sub-model, and let gl be an attention mechanism function. Let θ be the text feature vector of the resource text of the first object's historical access resource at the (M-1)th layer of the initial transformer submodel, where θ is the reference matrix used to represent... and The attention weights between them, where t is a positive integer less than N and greater than 1.

[0147] The processes for determining the resource identifier of the first object's historical access resource in the Mth layer of the initial transformer sub-model, the process for determining the image feature vector of the resource image of the first object's historical access resource in the Mth layer of the initial transformer sub-model, the process for determining the text feature vector of the resource text of the third object's historical access resource in the Mth layer of the initial transformer sub-model, the process for determining the resource identifier of the third object's historical access resource in the Mth layer of the initial transformer sub-model, and the process for determining the image feature vector of the resource image of the third object's historical access resource in the Mth layer of the initial transformer sub-model are similar to the process for determining the text feature vector of the resource text of the first object's historical access resource in the Mth layer of the initial transformer sub-model, and will not be repeated here.

[0148] like Figure 7 This is a flowchart illustrating how to determine a second target feature vector according to an embodiment of this application. Based on historical access data of a first object, text feature vectors, image feature vectors, and identifier feature vectors of the first object at the first layer are obtained. Then, these vectors are updated at layers 2 through M to obtain the text feature vectors, image feature vectors, and identifier feature vectors of the first object at layer M. Based on these vectors, the feature vector of the first object at layer M is obtained. Similarly, based on historical access data of a third object, text feature vectors, image feature vectors, and identifier feature vectors of the third object at the first layer are obtained. Then, these vectors are updated at layers 2 through M to obtain the text feature vectors, image feature vectors, and identifier feature vectors of the third object at layer M. Based on these vectors, the feature vector of the third object at layer M is obtained. The feature vectors of the first object at layer M and the feature vectors of the third object at layer M are fused to obtain a fused feature vector. The fused feature vector is then updated through layers M+1 to N to obtain the second target feature vector.

[0149] In one possible implementation, after determining the first target feature vector and the second target feature vector, the target loss value can be determined in the following two ways based on the first target feature vector and the second target feature vector.

[0150] Implementation Method 1: Determine the first loss value based on the first target feature vector and the second target feature vector, and use the first loss value as the target loss value.

[0151] The first loss value is determined according to the first target feature vector and the second target feature vector, according to the following formula (5).

[0152]

[0153] In the above formula (5), L cl Let be the first loss value, and log be the logarithmic function. The first target feature vector, v is the feature vector of the second target. u- Let E be the feature vector of the second object, and let E be a set including multiple second objects.

[0154] Method 2: Determine the first loss value based on the first target feature vector and the second target feature vector; determine the second loss value based on the first object feature vector, the resource feature vector of the first resource, and the resource feature vector of the second resource; and determine the target loss value based on the first loss value and the second loss value.

[0155] Here, the first resource refers to the resources that the first object has historically accessed, and the second resource refers to the resources that the first object has not accessed. The process of determining the first loss value based on the first target feature vector and the second target feature vector has been described in the above implementation method one, and will not be repeated here.

[0156] In one possible implementation, the second loss value is determined according to the following formula (6) based on the first object feature vector, the resource feature vector of the first resource, and the resource feature vector of the second resource.

[0157]

[0158] In the above formula (6), L main The second loss value is given by log, where log is the logarithmic function, and v u The first object's feature vector. This is the resource feature vector of the first resource. Let I be the resource feature vector of the second resource. - This is a collection containing resources that the first object has not accessed. The collection containing resources that the first object has not accessed includes at least one resource.

[0159] Optionally, after determining the first loss value and the second loss value, the target loss value is determined according to the first loss value and the second loss value in accordance with the following formula (7).

[0160] L total=L main +β·L cl Formula (7)

[0161] In the above formula (7), L total L is the target loss value. cl L is the first loss value. main β is the second loss value, and β is the proportion of the first loss value. The proportion of the first loss value is set based on experience or adjusted according to the implementation environment, and this application embodiment does not limit it in this way.

[0162] It should be noted that the target loss value can be determined according to any of the above implementation methods, and the embodiments of this application do not limit this.

[0163] In step 205, the initial resource push model is updated based on the target loss value being greater than the loss value threshold, to obtain the target resource push model.

[0164] In one possible implementation, after determining the target loss value, the initial resource push model is updated based on the target loss value being greater than a loss value threshold, resulting in the target resource push model. Since the initial resource push model includes at least one of an initial text vector determination sub-model, an initial identifier vector determination sub-model, an initial image vector determination sub-model, and an initial transformer sub-model, the process of updating the initial resource push model includes, but is not limited to, updating at least one of the following: updating the parameters of the initial text vector determination sub-model, updating the parameters of the initial identifier vector determination sub-model, updating the parameters of the initial image vector determination sub-model, and updating the parameters of the initial transformer sub-model. The loss value threshold is set based on experience or adjusted according to the implementation environment; this embodiment does not limit its application.

[0165] The initial resource push model is updated to obtain a first resource push model. Steps 201 to 204 above are then executed using the first resource push model to obtain a loss value. If the loss value is not greater than a loss value threshold, the first resource push model is used as the target resource push model. If the loss value is greater than the loss value threshold, the first resource push model continues to be updated until the loss value obtained by calling the updated resource push model is not greater than the loss value threshold. At this point, the updated resource push model is used as the target resource push model.

[0166] Optionally, if the target loss value is not greater than the loss value threshold, the initial resource push model is used as the target resource push model.

[0167] The resource push model acquisition method provided in this application has achieved significant improvements in Recall, MRR (Mean reciprocal rank, a mechanism for evaluating search algorithms), NDCG (Normalized Discounted Cumulative Gain), CTR (Click Through Rate), CVR (Conversion Rate), and IPAY (Resource Transfer Rate). Specifically, it improves Recall by approximately 2%-6%, MRR by approximately 10%-14%, and NDCG by approximately 10%-13%. CTR is improved by 1.212%, CVR by 1.203%, and IPAY by 2.362%.

[0168] The above method, when updating the initial resource push model, considers not only the historical access data of the first object but also the historical access data of the second object. By identifying related objects in the second object that are relevant to the first object, the initial resource push model is updated to ensure that the accuracy of the updated target resource push model is high. This allows for more accurate resource push to the object, resulting in a higher matching degree between the pushed resource and the object, increasing the probability that the pushed resource will be clicked by the object, and thus improving the effect of resource push.

[0169] Figure 8 This is a flowchart of a resource push method provided in an embodiment of this application. The method can be... Figure 1 The electronic device in the process can be executed by the electronic device. Figure 1 Terminal device 101 in the middle can also be Figure 1 The server 102 in this embodiment is not limited thereto. Figure 8 As shown, the method includes the following steps 801 to 803.

[0170] In step 801, the historical access data of the target object and the target resource push model are obtained.

[0171] The target resource push model is based on the above. Figure 2The method embodiment shown is obtained. The historical access data of the target object includes, but is not limited to, at least one of the following: resource text, resource identifier, and resource image of the resources historically accessed by the target object. The resource text describes the resource; the resource identifier can be a resource number, a resource name, or any identifier that can represent the resource (this embodiment does not limit this); and the resource image is an image of the resource. The historical access resources of the target object include, but are not limited to, at least one of the following: resources historically browsed by the target object, resources historically clicked by the target object, resources historically converted by the target object, and resources historically searched by the target object. For example, the historical access data of the target object includes: product B, the descriptive text of product B, and an image of product B.

[0172] Optionally, the electronic device is a terminal device, which has an application installed and running for acquiring resources. This application can acquire resources of any type, and this embodiment of the application does not limit this. The terminal device's display interface displays relevant information about the application, which may include the application's icon. When a target object wants to acquire resources, the target object selects the application's icon, the terminal device receives the target object's selection instruction for the application's icon, and the terminal device acquires the target object's historical access data.

[0173] When an electronic device acts as a server, upon receiving a selection instruction from a target object for an application icon, the terminal device sends a resource recommendation request to the server. This request carries the object identifier of the target object. Upon receiving the resource recommendation request, the server parses it to obtain the object identifier and then uses that identifier to retrieve the target object's historical access data.

[0174] In step 802, the target resource push model is invoked to process the historical access data of the target object, resulting in multiple target resources.

[0175] In one possible implementation, the target resource push model is invoked to process the historical access data of the target object, obtaining the object feature vector of the target object. This object feature vector is used to characterize the target object. Based on the object feature vector of the target object and the resource feature vectors of multiple candidate resources, multiple target resources are determined from among the candidate resources.

[0176] The electronic device stores resource feature vectors of multiple candidate resources. The process of determining multiple target resources from among the candidate resources based on the object feature vector of the target object and the resource feature vectors of the multiple candidate resources includes: determining the matching degree between each candidate resource and the target object based on the object feature vector of the target object and the resource feature vectors of the multiple candidate resources; and selecting candidate resources whose matching degree meets the matching requirements as target resources. Here, "meeting the matching requirements" means that the matching degree is greater than a matching threshold. The matching threshold is set based on experience or adjusted according to the implementation environment; this embodiment does not limit this.

[0177] Optionally, the matching degree between each candidate resource and the target object is determined according to the object feature vector of the target object and the resource feature vectors of multiple candidate resources, according to the following formula (8).

[0178] S i =P·Q i Formula (8)

[0179] In the above formula (8), S i Let P be the matching degree between the i-th candidate resource and the target object, and Q be the object feature vector of the target object. i Let be the resource feature vector of the i-th candidate resource.

[0180] In step 803, multiple target resources are pushed to the target object.

[0181] After identifying multiple target resources in step 802 above, the multiple target resources are pushed to the target object. Optionally, if the electronic device is a terminal device, after identifying multiple target resources in step 802 above, the terminal device displays the resource information of the multiple target resources to achieve the purpose of pushing multiple target resources to the target object. If the electronic device is a server, after identifying multiple target resources in step 802 above, the server sends the resource information of the multiple target resources to the terminal device, and the terminal device displays the resource information of the multiple target resources to achieve the purpose of pushing multiple target resources to the target object.

[0182] The above method uses a highly accurate target resource push model to push resources to target objects, which makes the target resources highly matched with the target objects, increases the probability that the target resources will be clicked by the target objects, and thus improves the effect of resource push.

[0183] Figure 9 The diagram shown is a structural schematic of a resource push model acquisition device provided in an embodiment of this application. Figure 9 As shown, the device includes:

[0184] The processing module 901 is used to call the initial resource push model to process the historical access data of the first object and obtain the feature vector of the first object. The feature vector of the first object is used to characterize the first object.

[0185] The processing module 901 is also used to call the initial resource push model to process the historical access data of multiple second objects to obtain multiple feature vectors of second objects, and any feature vector of a second object is used to characterize any second object.

[0186] The determining module 902 is used to determine at least one related object among the multiple second objects based on the feature vector of the first object and the feature vectors of multiple second objects. The related object is an object whose correlation with the first object meets the correlation requirements.

[0187] The determination module 902 is further configured to determine a target loss value based on the feature vector of the first object and the feature vector of the second object of at least one related object, the target loss value being used to indicate the accuracy of the initial resource push model;

[0188] The update module 903 is used to update the initial resource push model based on the target loss value being greater than the loss value threshold, so as to obtain the target resource push model.

[0189] In one possible implementation, the processing module 901 is used to process the historical access data of the first object by calling the initial resource push model, based on the fact that the historical access data of the first object is one, to obtain the feature vector of the historical access data of the first object, and to use the feature vector of the historical access data of the first object as the feature vector of the first object; or, based on the fact that the historical access data of the first object is multiple, to determine the target historical access data from the multiple historical access data of the first object, to process the target historical access data by calling the initial resource push model, to obtain the feature vector of the target historical access data, and to use the feature vector of the target historical access data as the feature vector of the first object.

[0190] In one possible implementation, the processing module 901 is used to determine the feature vector of the historical access data of the first object, which includes the resource text, resource identifier, and resource image of the historical accessed resources of the first object. The initial resource push model includes an initial text vector determination sub-model, an initial identifier vector determination sub-model, and an initial image vector determination sub-model. The initial text vector determination sub-model is called to process the resource text of the historical accessed resources of the first object to obtain the text feature vector of the historical accessed resources of the first object; the initial image vector determination sub-model is called to process the resource image of the historical accessed resources of the first object to obtain the image feature vector of the historical accessed resources of the first object; the initial identifier vector determination sub-model is called to process the resource identifier, text feature vector, and image feature vector of the historical accessed resources of the first object to obtain the identifier feature vector of the historical accessed resources of the first object; and the feature vector of the historical access data of the first object is determined based on the text feature vector, image feature vector, and identifier feature vector of the historical accessed resources of the first object.

[0191] In one possible implementation, the processing module 901 is used to obtain the access time of multiple historical access data of the first object; and to select the historical access data whose access time meets the time requirement from the multiple historical access data of the first object as the target historical access data.

[0192] In one possible implementation, the determining module 902 is used to determine the similarity between each second object and the first object based on the feature vector of the first object and the feature vectors of multiple second objects; and to take the second objects whose similarity meets the similarity requirements as related objects.

[0193] In one possible implementation, the determining module 902 is configured to: determine the weights of each related object based on the feature vector of a first object and the feature vectors of at least one related object of a second object, wherein the weight of any related object is used to indicate the influence of any related object on the first object; determine a first target feature vector based on the feature vectors of each related object of the second object and the weights of each related object, wherein the first target feature vector is used to characterize the interest preferences of the first object; determine a second target feature vector based on the historical access data of the first object and the historical access data of a third object, wherein the second target feature vector is used to characterize the interest preferences of the first object, and the third object is any one of the at least one related object; and determine a target loss value based on the first target feature vector and the second target feature vector.

[0194] In one possible implementation, the determining module 902 is further configured to determine a second loss value based on the feature vector of the first object, the resource feature vector of the first resource, and the resource feature vector of the second resource, wherein the first resource is a resource that the first object has historically accessed, and the second resource is a resource that the first object has not accessed.

[0195] The determination module 902 is used to determine a first loss value based on a first target feature vector and a second target feature vector; and to determine a target loss value based on the first loss value and the second loss value.

[0196] In one possible implementation, the initial resource push model includes an initial transformer sub-model, which has N layers, where N is a positive integer.

[0197] The determination module 902 is used to determine the feature vector of the historical access data of the first object at the Mth layer of the initial converter sub-model based on the historical access data of the first object, including the resource text, resource identifier, and resource image of the historical accessed resources of the first object, where M is greater than 0 and less than N; based on the historical access data of the third object, including the resource text, resource identifier, and resource image of the historical accessed resources of the third object, the feature vector of the historical access data of the third object at the Mth layer of the initial converter sub-model is determined based on the resource text, resource identifier, and resource image of the historical accessed resources of the third object; based on the feature vector of the historical access data of the first object at the Mth layer of the initial converter sub-model and the feature vector of the historical access data of the third object at the Mth layer of the initial converter sub-model, a fused feature vector is determined; and the fused feature vector is updated through the (M+1)th to the Nth layers of the initial converter sub-model to obtain the second target feature vector.

[0198] In one possible implementation, the determining module 902 is used to determine, based on the resource text, resource identifier, and resource image of the historically accessed resources of the first object, the reference text feature vector of the resource text of the historically accessed resources of the first object at the M-th layer of the initial converter sub-model, the reference identifier feature vector of the resource identifier of the historically accessed resources of the first object at the M-th layer of the initial converter sub-model, and the reference image feature vector of the resource image of the historically accessed resources of the first object at the M-th layer of the initial converter sub-model; and to fuse the reference text feature vector, the reference identifier feature vector, and the reference image feature vector to obtain the feature vector of the historical access data of the first object at the M-th layer of the initial converter sub-model.

[0199] When updating the initial resource push model, the device considers not only the historical access data of the first object but also the historical access data of the second object. By identifying related objects in the second object that are relevant to the first object, the device updates the initial resource push model to ensure that the accuracy of the updated target resource push model is high. This allows for more accurate resource push to the object, resulting in a higher matching degree between the pushed resource and the object, increasing the probability that the pushed resource will be clicked by the object, and thus improving the effect of resource push.

[0200] Figure 10 This is a schematic diagram of the structure of a resource push device provided in an embodiment of this application, as shown below. Figure 10 As shown, the device includes:

[0201] Module 1001 is used to acquire historical access data of the target object and the target resource push model, the target resource push model being based on the above. Figure 9 The apparatus shown is used to obtain;

[0202] Processing module 1002 is used to call the target resource push model to process the historical access data of the target object and obtain multiple target resources;

[0203] The push module 1003 is used to push multiple target resources to the target object.

[0204] In one possible implementation, the processing module 1002 is used to call the target resource push model to process the historical access data of the target object to obtain the object feature vector of the target object, which is used to characterize the target object; based on the object feature vector of the target object and the resource feature vectors of multiple candidate resources, multiple target resources are determined from the multiple candidate resources.

[0205] The aforementioned device employs a highly accurate target resource push model to push resources to target objects, resulting in a high degree of matching between the pushed target resources and the target objects. This increases the probability that the target resources will be clicked by the target objects, thereby improving the effectiveness of resource push.

[0206] It should be understood that the above-described apparatus is only illustrated by the division of the functional modules described above when implementing its functions. In practical applications, the functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process can be found in the method embodiments, which will not be repeated here.

[0207] Figure 11A structural block diagram of a terminal device 1100 provided in an exemplary embodiment of this application is shown. The terminal device 1100 may be a portable mobile terminal, such as a smartphone, tablet computer, MP3 player (Moving Picture Experts Group Audio Layer III), MP4 player (Moving Picture Experts Group Audio Layer IV), laptop computer, or desktop computer. The terminal device 1100 may also be referred to as a user device, portable terminal, laptop terminal, desktop terminal, or other names.

[0208] Typically, terminal device 1100 includes a processor 1101 and a memory 1102.

[0209] Processor 1101 may include one or more processing cores, such as a quad-core processor, an octa-core processor, etc. Processor 1101 may be implemented using at least one hardware form selected from DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). Processor 1101 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, processor 1101 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content required to be displayed on the screen. In some embodiments, processor 1101 may also include an AI (Artificial Intelligence) processor, which is used to handle computational operations related to machine learning.

[0210] The memory 1102 may include one or more computer-readable storage media, which may be non-transitory. The memory 1102 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments, the non-transitory computer-readable storage media in the memory 1102 is used to store at least one instruction, which is executed by the processor 1101 to implement the present application. Figure 2The method embodiment shown provides a method for obtaining a resource push model, or, to implement the method in this application. Figure 8 The resource push method provided in the illustrated embodiment.

[0211] In some embodiments, the terminal device 1100 may also optionally include: a peripheral device interface 1103 and at least one peripheral device. The processor 1101, memory 1102, and peripheral device interface 1103 can be connected via a bus or signal line. Each peripheral device can be connected to the peripheral device interface 1103 via a bus, signal line, or circuit board. Specifically, the peripheral device includes at least one of: a radio frequency circuit 1104, a display screen 1105, a camera assembly 1106, an audio circuit 1107, a positioning assembly 1108, and a power supply 1109.

[0212] Peripheral device interface 1103 can be used to connect at least one I / O (Input / Output) related peripheral device to processor 1101 and memory 1102. In some embodiments, processor 1101, memory 1102 and peripheral device interface 1103 are integrated on the same chip or circuit board; in some other embodiments, any one or two of processor 1101, memory 1102 and peripheral device interface 1103 can be implemented on separate chips or circuit boards, which is not limited in this embodiment.

[0213] Display screen 1105 is used to display a UI (User Interface). This UI may include graphics, text, icons, videos, and any combination thereof. When display screen 1105 is a touch display screen, it also has the ability to collect touch signals on or above its surface. These touch signals can be input as control signals to processor 1101 for processing. In this case, display screen 1105 can also be used to provide virtual buttons and / or a virtual keyboard, also known as soft buttons and / or a soft keyboard. In some embodiments, there may be one display screen 1105, disposed on the front panel of terminal device 1100; in other embodiments, there may be at least two display screens, disposed on different surfaces of terminal device 1100 or in a folded design; in still other embodiments, display screen 1105 may be a flexible display screen, disposed on a curved or folded surface of terminal device 1100. Furthermore, display screen 1105 may be configured as a non-rectangular, irregular shape, i.e., a non-rectangular screen. The display screen 1105 can be made of materials such as LCD (Liquid Crystal Display) and OLED (Organic Light-Emitting Diode).

[0214] The camera assembly 1106 is used to acquire images or videos. Optionally, the camera assembly 1106 includes a front-facing camera and a rear-facing camera. Typically, the front-facing camera is located on the front panel of the terminal device 1100, and the rear-facing camera is located on the back of the terminal device 1100. In some embodiments, there are at least two rear-facing cameras, which are any one of a main camera, a depth-sensing camera, a wide-angle camera, and a telephoto camera, to achieve background blurring by fusion of the main camera and the depth-sensing camera, panoramic shooting by fusion of the main camera and the wide-angle camera, VR (Virtual Reality) shooting, or other fusion shooting functions. In some embodiments, the camera assembly 1106 may also include a flash. The flash can be a single-color temperature flash or a dual-color temperature flash. A dual-color temperature flash refers to a combination of a warm light flash and a cool light flash, which can be used for light compensation at different color temperatures.

[0215] The audio circuit 1107 may include a microphone and a speaker. The microphone is used to collect sound waves from the user and the environment, converting the sound waves into electrical signals that are input to the processor 1101 for processing, or input to the radio frequency circuit 1104 for voice communication. For stereo sound acquisition or noise reduction purposes, multiple microphones may be used, each located at a different part of the terminal device 1100. The microphone may also be an array microphone or an omnidirectional microphone. The speaker is used to convert electrical signals from the processor 1101 or the radio frequency circuit 1104 into sound waves. The speaker may be a conventional diaphragm speaker or a piezoelectric ceramic speaker. When the speaker is a piezoelectric ceramic speaker, it can convert electrical signals not only into audible sound waves but also into inaudible sound waves for purposes such as distance measurement. In some embodiments, the audio circuit 1107 may also include a headphone jack.

[0216] The positioning component 1108 is used to locate the current geographical location of the terminal device 1100 in order to enable navigation or LBS (Location Based Service). The positioning component 1108 can be a positioning component based on the US GPS (Global Positioning System), China's BeiDou system, Russia's Granas system, or the European Union's Galileo system.

[0217] Power supply 1109 is used to supply power to the various components in terminal device 1100. Power supply 1109 can be AC ​​power, DC power, a disposable battery, or a rechargeable battery. When power supply 1109 includes a rechargeable battery, the rechargeable battery can be a wired rechargeable battery or a wireless rechargeable battery. A wired rechargeable battery is a battery that is charged via a wired line, and a wireless rechargeable battery is a battery that is charged via a wireless coil. The rechargeable battery can also be used to support fast charging technology.

[0218] In some embodiments, the terminal device 1100 further includes one or more sensors 1110. The one or more sensors 1110 include, but are not limited to: an accelerometer 1111, a gyroscope 1112, a pressure sensor 1113, a fingerprint sensor 1114, an optical sensor 1115, and a proximity sensor 1116.

[0219] Accelerometer 1111 can detect the magnitude of acceleration along the three coordinate axes of a coordinate system established by terminal device 1100. For example, accelerometer 1111 can be used to detect the components of gravitational acceleration along the three coordinate axes. Processor 1101 can control display screen 1105 to display the user interface in either a landscape or portrait view based on the gravitational acceleration signal acquired by accelerometer 1111. Accelerometer 1111 can also be used for games or for acquiring user motion data.

[0220] The gyroscope sensor 1112 can detect the orientation and rotation angle of the terminal device 1100. The gyroscope sensor 1112 can work in conjunction with the accelerometer sensor 1111 to collect the user's 3D movements on the terminal device 1100. Based on the data collected by the gyroscope sensor 1112, the processor 1101 can perform the following functions: motion sensing (e.g., changing the UI based on the user's tilt), image stabilization during shooting, game control, and inertial navigation.

[0221] The pressure sensor 1113 can be disposed on the side bezel of the terminal device 1100 and / or on the lower layer of the display screen 1105. When the pressure sensor 1113 is disposed on the side bezel of the terminal device 1100, it can detect the user's grip signal on the terminal device 1100, and the processor 1101 can perform left / right hand recognition or quick operation based on the grip signal collected by the pressure sensor 1113. When the pressure sensor 1113 is disposed on the lower layer of the display screen 1105, the processor 1101 can control the operable controls on the UI interface based on the user's pressure operation on the display screen 1105. The operable controls include at least one of button controls, scroll bar controls, icon controls, and menu controls.

[0222] The fingerprint sensor 1114 is used to collect the user's fingerprint. The processor 1101 identifies the user's identity based on the fingerprint collected by the fingerprint sensor 1114, or the fingerprint sensor 1114 identifies the user's identity based on the collected fingerprint. When the user's identity is identified as trusted, the processor 1101 authorizes the user to perform relevant sensitive operations, including unlocking the screen, viewing encrypted information, downloading software, making payments, and changing settings. The fingerprint sensor 1114 can be located on the front, back, or side of the terminal device 1100. When the terminal device 1100 has a physical button or manufacturer logo, the fingerprint sensor 1114 can be integrated with the physical button or manufacturer logo.

[0223] An optical sensor 1115 is used to collect ambient light intensity. In one embodiment, the processor 1101 can control the display brightness of the display screen 1105 based on the ambient light intensity collected by the optical sensor 1115. Specifically, when the ambient light intensity is high, the display brightness of the display screen 1105 is increased; when the ambient light intensity is low, the display brightness of the display screen 1105 is decreased. In another embodiment, the processor 1101 can also dynamically adjust the shooting parameters of the camera assembly 1106 based on the ambient light intensity collected by the optical sensor 1115.

[0224] The proximity sensor 1116, also known as a distance sensor, is typically located on the front panel of the terminal device 1100. The proximity sensor 1116 is used to detect the distance between the user and the front of the terminal device 1100. In one embodiment, when the proximity sensor 1116 detects that the distance between the user and the front of the terminal device 1100 is gradually decreasing, the processor 1101 controls the display screen 1105 to switch from a screen-on state to a screen-off state; when the proximity sensor 1116 detects that the distance between the user and the front of the terminal device 1100 is gradually increasing, the processor 1101 controls the display screen 1105 to switch from a screen-off state to a screen-on state.

[0225] Those skilled in the art will understand that Figure 11 The structure shown does not constitute a limitation on the terminal device 1100, and may include more or fewer components than shown, or combine certain components, or use different component arrangements.

[0226] Figure 12This is a schematic diagram of the server structure provided in the embodiments of this application. The server 1200 can vary considerably due to different configurations or performance. It may include one or more Central Processing Units (CPUs) 1201 and one or more memories 1202, wherein the one or more memories 1202 store at least one line of program code, which is loaded and executed by the one or more processors 1201 to implement the above-mentioned... Figure 2 The method embodiment shown provides a method for obtaining the resource push model, or, to implement the above... Figure 8 The resource push method provided in the illustrated embodiment is as follows. Of course, the server 1200 may also have wired or wireless network interfaces, a keyboard, and input / output interfaces for input / output. The server 1200 may also include other components for implementing device functions, which will not be elaborated upon here.

[0227] In an exemplary embodiment, a computer-readable storage medium is also provided, which stores at least one line of program code, which is loaded and executed by a processor to enable the computer to perform the above-described functions. Figure 2 The method embodiment shown provides a method for obtaining the resource push model, or implements the above. Figure 8 The resource push method provided in the illustrated embodiment.

[0228] Optionally, the aforementioned computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, floppy disk, and optical data storage device, etc.

[0229] In an exemplary embodiment, a computer program or computer program product is also provided, which stores at least one computer instruction, which is loaded and executed by a processor to enable the computer to perform the above-described functions. Figure 2 The method embodiment shown provides a method for obtaining the resource push model, or implements the above. Figure 8 The resource push method provided in the illustrated embodiment.

[0230] It should be noted that all information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.), and signals involved in this application have been authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions. For example, the historical access data involved in this application was obtained with full authorization.

[0231] It should be understood that "multiple" as used in this article refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0232] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0233] The above description is merely an exemplary embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the principles of this application should be included within the protection scope of this application.

Claims

1. A method for obtaining a resource push model, characterized in that, The method includes: The initial resource push model is invoked to process the historical access data of the first object to obtain the feature vector of the first object, which is used to characterize the first object. The initial resource push model is invoked to process the historical access data of multiple second objects, resulting in multiple feature vectors of the second objects. Any feature vector of a second object is used to characterize any second object. Based on the first object feature vector and the plurality of second object feature vectors, at least one related object is determined among the plurality of second objects, wherein the related object is an object whose correlation with the first object meets the correlation requirements; A target loss value is determined based on the feature vector of the first object and the feature vector of the second object of the at least one related object, the target loss value being used to indicate the accuracy of the initial resource push model; Based on the fact that the target loss value is greater than the loss value threshold, the initial resource push model is updated to obtain the target resource push model; Determining the target loss value based on the first object feature vector and the second object feature vector of the at least one related object includes: Based on the feature vector of the first object and the feature vector of the second object of the at least one related object, the weight of each related object is determined, and the weight of any related object is used to indicate the influence of any related object on the first object; Based on the second object feature vectors of each related object and the weights of each related object, a first target feature vector is determined, which is used to characterize the interest preferences of the first object. Based on the historical access data of the first object and the historical access data of the third object, a second target feature vector is determined. The second target feature vector is used to characterize the interest preferences of the first object. The third object is any one of the at least one related object. The target loss value is determined based on the first target feature vector and the second target feature vector.

2. The method according to claim 1, characterized in that, The process of calling the initial resource push model to process the historical access data of the first object to obtain the feature vector of the first object includes: Based on the fact that the historical access data of the first object is one, the initial resource push model is called to process the historical access data of the first object to obtain the feature vector of the historical access data of the first object, and the feature vector of the historical access data of the first object is used as the feature vector of the first object. Since there are multiple historical access data for the first object, target historical access data is determined from the multiple historical access data of the first object. The initial resource push model is called to process the target historical access data to obtain the feature vector of the target historical access data. The feature vector of the target historical access data is used as the feature vector of the first object.

3. The method according to claim 2, characterized in that, The step of calling the initial resource push model to process the historical access data of the first object to obtain the feature vector of the historical access data of the first object includes: Based on the historical access data of the first object, including the resource text, resource identifier, and resource image of the historical access resources of the first object, the initial resource push model includes an initial text vector determination sub-model, an initial identifier vector determination sub-model, and an initial image vector determination sub-model; The initial text vector determination sub-model is invoked to process the resource text of the first object's historical access resources, thereby obtaining the text feature vector of the first object's historical access resources; The initial image vector determination sub-model is called to process the resource images of the historical accessed resources of the first object to obtain the image feature vector of the historical accessed resources of the first object; The initial identifier vector determination sub-model is invoked to process the resource identifier of the historical accessed resources of the first object, the text feature vector, and the image feature vector to obtain the identifier feature vector of the historical accessed resources of the first object; Based on the text feature vector, image feature vector, and identifier feature vector of the historical access resources of the first object, the feature vector of the historical access data of the first object is determined.

4. The method according to claim 2, characterized in that, The step of determining the target historical access data from multiple historical access data of the first object includes: Obtain the access times of multiple historical access data for the first object; The historical access data that meets the time requirement among the multiple historical access data of the first object is taken as the target historical access data.

5. The method according to any one of claims 1 to 4, characterized in that, The step of determining at least one related object among the plurality of second objects based on the first object feature vector and the plurality of second object feature vectors includes: Based on the feature vector of the first object and the feature vectors of the plurality of second objects, the similarity between each second object and the first object is determined. The second object whose similarity meets the similarity requirement is considered as the related object.

6. The method according to claim 1, characterized in that, The method further includes: A second loss value is determined based on the feature vector of the first object, the resource feature vector of the first resource, and the resource feature vector of the second resource. The first resource is the resource that the first object has historically accessed, and the second resource is the resource that the first object has not accessed. Determining the target loss value based on the first target feature vector and the second target feature vector includes: Based on the first target feature vector and the second target feature vector, determine the first loss value; The target loss value is determined based on the first loss value and the second loss value.

7. A resource push method, characterized in that, The method includes: Obtain historical access data of the target object and a target resource push model, wherein the target resource push model is obtained based on the method described in any one of claims 1 to 6 above; The target resource push model is invoked to process the historical access data of the target object, resulting in multiple target resources; The multiple target resources are pushed to the target object.

8. The method according to claim 7, characterized in that, The process of calling the target resource push model to process the historical access data of the target object yields multiple target resources, including: The target resource push model is invoked to process the historical access data of the target object to obtain the object feature vector of the target object, which is used to characterize the target object. Based on the object feature vector of the target object and the resource feature vectors of multiple candidate resources, multiple target resources are determined from the multiple candidate resources.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one piece of program code, which is loaded and executed by a processor to enable the computer to implement the resource push model acquisition method as described in any one of claims 1 to 6, or to enable the computer to implement the resource push method as described in claim 7 or 8.

Images