CN116628309A

CN116628309A - Resource recommendation method and device, electronic equipment and storage medium

Info

Publication number: CN116628309A
Application number: CN202210125372.1A
Authority: CN
Inventors: 汪的; 万熙颖
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2022-02-10
Filing date: 2022-02-10
Publication date: 2023-08-22

Abstract

The application discloses a resource recommendation method, a device, electronic equipment and a storage medium, wherein the method comprises the following steps: and constructing and obtaining recommended relation information by taking target recommended index data, object associated data and resource associated data as nodes and taking the associated relation among the data corresponding to the nodes as edges, inputting historical object resource data and recommended relation information into an influence factor analysis model for data influence analysis, obtaining an influence factor analysis result, representing influence distribution information of the historical object resource data on the target recommended index data by the influence factor analysis result, and executing resource recommendation processing based on the influence factor analysis result. The method quantifies the influence of each influence data on the target recommendation index data, improves the accuracy and the effectiveness of influence factor analysis, enables the influence factor analysis result to be applied to the resource recommendation service scene, and improves the rationality of the resource recommendation service.

Description

Resource recommendation method and device, electronic equipment and storage medium

Technical Field

The present application relates to the field of recommendation technologies, and in particular, to a resource recommendation method, device, electronic device, and storage medium.

Background

The user experience evaluation is a technology capable of improving functions in a product by aiming at operation data of the product used by a user after the product is online, in the prior art, a user experience analysis framework is index information, connection from the user experience data to integral data of the product application cannot be provided, and the corresponding quantity of the index information is often larger than user experience data summarized by actual product business, so that a result obtained after the user experience is analyzed is more ideal, and accuracy and effectiveness are low when impact factor analysis is carried out on target recommendation index data.

Disclosure of Invention

The application provides a resource recommendation method, a resource recommendation device, electronic equipment and a storage medium, which can improve the accuracy and the effectiveness of influence factor analysis.

In one aspect, the present application provides a resource recommendation method, which includes:

acquiring historical object resource data corresponding to a target object, wherein the historical object resource data comprises object associated data of the target object and resource associated data of historical multimedia resources recommended to the target object;

the target recommendation index data, the object association data and the resource association data are taken as nodes, and association relations among the data corresponding to the nodes are taken as edges to construct recommendation relation information;

Inputting the historical object resource data and the recommendation relation information into an influence factor analysis model for data influence analysis to obtain an influence factor analysis result, wherein the influence factor analysis result represents influence distribution information of the historical object resource data on the target recommendation index data;

and executing resource recommendation processing based on the influence factor analysis result.

Another aspect provides a resource recommendation apparatus, the apparatus comprising:

the historical data acquisition module is used for acquiring historical object resource data corresponding to a target object, wherein the historical object resource data comprises object associated data of the target object and resource associated data of historical multimedia resources recommended to the target object;

the recommendation relation information construction module is used for constructing recommendation relation information by taking target recommendation index data, the object association data and the resource association data as nodes and taking association relations among the data corresponding to the nodes as edges;

the data influence analysis module is used for inputting the historical object resource data and the recommendation relation information into an influence factor analysis model to perform data influence analysis to obtain an influence factor analysis result, and the influence factor analysis result represents influence distribution information of the historical object resource data on the target recommendation index data;

And the resource recommendation module is used for executing resource recommendation processing based on the influence factor analysis result.

In another aspect, an electronic device is provided, the electronic device including a processor and a memory, the memory storing at least one instruction or at least one program, the at least one instruction or the at least one program loaded and executed by the processor to implement a resource recommendation method as described above.

Another aspect provides a computer readable storage medium comprising a processor and a memory having stored therein at least one instruction or at least one program loaded and executed by the processor to implement a resource recommendation method as described above.

In another aspect, a computer program product is provided, comprising a computer program, which when executed by a processor implements the resource recommendation method described above.

According to the resource recommendation method, the device, the electronic equipment and the storage medium, target recommendation index data, object association data and resource association data are taken as nodes, association relations among the corresponding data of the nodes are taken as edges to construct recommendation relation information, historical object resource data and recommendation relation information are input into an influence factor analysis model to conduct data influence analysis, an influence factor analysis result is obtained, the influence factor analysis result represents influence distribution information of the historical object resource data on the target recommendation index data, and resource recommendation processing is executed based on the influence factor analysis result. According to the method, the influence of each influence data on the target recommendation index data is quantized, so that recommendation analysis results can be returned to influence factor analysis results corresponding to different object groups, the accuracy and the effectiveness of influence factor analysis are improved, the influence factor analysis results can be applied to resource recommendation service scenes, and the rationality of resource recommendation services is improved.

Drawings

In order to more clearly illustrate the embodiments of the application or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 is an application scenario schematic diagram of a resource recommendation method provided in an embodiment of the present application;

FIG. 2 is a flowchart of a resource recommendation method according to an embodiment of the present application;

FIG. 3 is a flowchart of a method for constructing recommendation relationship information in a resource recommendation method according to an embodiment of the present application;

FIG. 4 is a schematic diagram of recommendation relationship information in a resource recommendation method according to an embodiment of the present application;

FIG. 5 is a flowchart of data impact analysis in a resource recommendation method according to an embodiment of the present application;

FIG. 6 is a schematic diagram corresponding to an analysis result of an influence factor when the duration of an advertisement video is 0-20s in the resource recommendation method according to the embodiment of the present application;

FIG. 7 is a flowchart of a method for recommending resources, according to an embodiment of the present application, for obtaining and fusing at least one model analysis result;

FIG. 8 is a flowchart of feature construction in a resource recommendation method according to an embodiment of the present application;

FIG. 9 is a schematic diagram of a feature construction manner in a resource recommendation method according to an embodiment of the present application;

FIG. 10 is a flowchart illustrating a resource recommendation operation performed based on the result of influence factor analysis in the resource recommendation method according to the embodiment of the present application;

FIG. 11 is a schematic diagram of a module for executing the resource recommendation method in the resource recommendation method according to the embodiment of the present application;

fig. 12 is a schematic structural diagram of a resource recommendation device according to an embodiment of the present application;

fig. 13 is a schematic hardware structure of an apparatus for implementing the method provided by the embodiment of the present application.

Detailed Description

The present application will be described in further detail with reference to the accompanying drawings, for the purpose of making the objects, technical solutions and advantages of the present application more apparent. It will be apparent that the described embodiments are only some, but not all, embodiments of the application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.

In the description of the present application, it should be understood that the terms "first," "second," and the like are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature. Moreover, the terms "first," "second," and the like, are used to distinguish between similar objects and do not necessarily describe a particular order or precedence. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the application described herein may be implemented in sequences other than those illustrated or otherwise described herein.

It will be appreciated that in the specific embodiments of the present application, related data such as user information is involved, and when the above embodiments of the present application are applied to specific products or technologies, user permissions or consents need to be obtained, and the collection, use and processing of related data need to comply with related laws and regulations and standards of related countries and regions.

Referring to fig. 1, an application scenario diagram of a resource recommendation method provided by an embodiment of the present application is shown, where the application scenario includes a client 110 and a server 120, and the server 120 obtains historical object resource data corresponding to the client 110, where the historical object resource data includes object association data of a target object and resource association data of a historical multimedia resource recommended to the target object. The server 120 uses the target recommendation index data, the object association data and the resource association data as nodes, uses the association relation between the corresponding data of the nodes as an edge to construct recommendation relation information, the server 120 inputs the historical object resource data and the recommendation relation information into an influence factor analysis model to perform data influence analysis, obtains an influence factor analysis result, and executes resource recommendation processing based on the influence factor analysis result.

In the embodiment of the present application, the client 110 includes a smart phone, a desktop computer, a tablet computer, a notebook computer, a digital assistant, a smart wearable device, an on-vehicle terminal, and other types of physical devices, and may also include software running in the physical devices, such as an application program, and the like. The operating system running on the entity device in the embodiment of the present application may include, but is not limited to, an android system, an IOS system, linux, unix, windows, and the like. The client 110 includes a UI (User Interface) layer, and the client 110 provides display of multimedia resources and collection of log data to the outside through the UI layer, and in addition, transmits history object resource data required for data analysis to the server 120 based on an API (Application Programming Interface, application program Interface).

In an embodiment of the present application, the server 120 may include a server that operates independently, or a distributed server, or a server cluster that is composed of a plurality of servers. The server 120 may include a network communication unit, a processor, a memory, and the like. Specifically, the server 120 may be configured to construct recommended relationship information, and input the historical object resource data and the recommended relationship information into an influence factor analysis model to perform data influence analysis, so as to obtain an influence factor analysis result.

Referring to fig. 2, a resource recommendation method is shown, which can be applied to a server side, and the method includes:

s210, acquiring historical object resource data corresponding to a target object, wherein the historical object resource data comprises object association data of the target object and resource association data of historical multimedia resources recommended to the target object;

in some embodiments, the object association data may be intrinsic association data of the target object, for example, age, sex, and the like of the target object, the resource association data may include interactive operation data of the target object on the historical multimedia resource and intrinsic association data of the historical multimedia resource, the interactive operation data may include playing time duration, playing integrity, praise, forwarding, sharing, commentary, and the like of the video, and the intrinsic association data may include category, content, and the like of the historical multimedia resource.

S220, constructing recommendation relation information by taking target recommendation index data, object association data and resource association data as nodes and taking association relations among the data corresponding to the nodes as edges;

in some embodiments, the association may be a causal relationship, where the causal relationship indicates that, in two nodes connected by a directional line in the recommendation relationship information, a node from which the directional line starts may generate a causal effect on a node reached by the directional line, that is, a change condition of the node from which the directional line starts may affect a change condition of the node reached by the directional line. The object association data and the resource association data may have a causal relationship, and the object association data and the target recommendation index data or the resource association data and the target recommendation index data may also have a causal relationship. Therefore, the object association data and the resource association data in the recommendation relationship information have direct or indirect causal relationship with the target recommendation index data.

In some embodiments, referring to fig. 3, constructing recommendation relationship information by using target recommendation index data, object association data and resource association data as nodes and association relationships among corresponding data of the nodes as edges includes:

s310, recommendation analysis is carried out on recommendation results corresponding to the historical object resource information, to-be-verified influence data associated with target recommendation index data is obtained, and the to-be-verified influence data comprises object associated data affecting the target recommendation index data and resource associated data affecting the target recommendation index data;

s320, determining a first association relationship between target recommendation index data and influence data to be verified;

s340, determining a second association relationship between the object association data and the resource association data;

s350, constructing recommendation relation information by taking target recommendation index data, object association data and resource association data as nodes and taking a first association relation and a second association relation as sides.

In some embodiments, when recommendation analysis is performed on a recommendation result corresponding to the historical object resource information, the historical object resource information in the target application and the historical object resource information in other applications except the target application in the same type of application can be compared transversely to obtain differences between the historical object resource information in the target application and the historical object resource information in other applications, so that to-be-verified influence data associated with the target recommendation index data is obtained through analysis of the differences.

For example, in the recommendation analysis of advertisement resources, differences between the exposure distribution of the application software 1 on the advertisement, the time length and the time sequence and the application software 1 and the application software 3 are analyzed. And acquiring the consumption distribution of the user of the three ends of the application software 1, the application software 2 and the application software 3, namely historical object resource data. And comparing the user consumption distribution of the three ends of the application software 1, the application software 2 and the application software 3 to obtain the difference information of the user consumption distribution, wherein the difference information comprises the knowledge class occupation ratio which is obviously higher than other classes in the overall class exposure distribution of the application software 1. Meanwhile, the application software 2 and the application software 3 are compared transversely, the difference of the product type exposure distribution of the application software 1 is obvious, and under certain product types, such as games, the occupation ratio of the application software 1 and the bid product is greatly different. In addition, in the overall advertisement duration exposure distribution of the application software 1, the advertisement duration of 0-20 s is obviously lower than that of the application software 2 and the application software 3, the advertisement duty ratio of 40-60 s is obviously higher than that of the application software 2 and the application software 3, and the advertisement duration exposure distribution of the application software 2 and the application software 3 is relatively similar. Therefore, the influence of factors such as advertisement categories, advertisement time length, distribution of the advertisement time length before and after the user consumes the video time sequence on the advertisement consumption condition of the user on the whole can be determined, namely the influence data to be verified, which are associated with the target recommendation index data, are obtained.

In some embodiments, there is a direct causal relationship between the to-be-verified impact data and the target recommendation index data, where the change condition of the to-be-verified impact data directly affects the change condition of the target recommendation index data, that is, the first association relationship between the target recommendation index data and the to-be-verified impact data. Because the second association relationship exists between the object association data and the resource association data, and the to-be-verified influence data is actually the object association data influencing the target recommendation index data and the resource association data influencing the target recommendation index data, other object association data or other resource association data except the to-be-verified influence data may have indirect causal relationship on the target recommendation index data, and the change condition of the target recommendation index data is indirectly influenced. In addition, in the case that the to-be-verified influence data includes a plurality of to-be-verified influence data, a second association relationship exists between every two to-be-verified influence data.

Therefore, the target recommendation index data, the object association data and the resource association data are taken as nodes, the first association relationship and the second association relationship are taken as edges, recommendation relationship information is constructed, the recommendation relationship information is in a causal relationship graph structure, and the graph comprises the object association data and the resource association data which have direct causal relationship and indirect causal relationship on the target recommendation index data. Nodes associated with business experience data, such as video feed distribution, distribution policies, etc., may also be added to the recommendation relationship information.

In some embodiments, please refer to fig. 4, which is a schematic diagram of the recommended relationship information shown in fig. 4. In the schematic diagram, the target recommendation index data is the advertisement average completion rate, the object association data comprises user distribution, and the resource association data comprises advertisement video duration and video content respectively. In addition, two nodes for distributing and distributing the strategy are added based on business experience data. Based on the schematic diagram, it can be determined that the advertisement video duration distribution and the user distribution are data which has direct influence on the average completion rate of the advertisement, and the user distribution has influence on the advertisement video duration distribution, that is to say, the user distribution is a common cause of the advertisement video duration distribution and the average completion rate of the advertisement, the advertisement video duration distribution cannot be controlled to be unchanged when the influence of the user distribution on the average completion rate of the advertisement is determined, otherwise, the influence degree of the distribution on the average completion rate of the advertisement is calculated inaccurately. In addition, the video content has an influence on the distribution of the advertisement video duration, that is, the video content is data indirectly influencing the uniform broadcast rate of the advertisement, and the video content also can interfere with the verification result when verifying the influence degree of the distribution of the advertisement video duration on the uniform broadcast rate of the advertisement. Therefore, the historical object resource data can be classified based on the recommendation relationship information, and the historical object resource data associated with the target recommendation index data can be determined.

And the recommendation relationship information is constructed through the causal relationship among the data corresponding to the nodes, and the common cause, the interference factors and the like can be determined based on the recommendation relationship information, so that the influence factor analysis model is convenient for carrying out data influence analysis, and the accuracy of the data influence analysis is improved.

S230, inputting the historical object resource data and the recommendation relation information into an influence factor analysis model to perform data influence analysis, so as to obtain an influence factor analysis result, wherein the influence factor analysis result represents influence distribution information of the historical object resource data on target recommendation index data;

in some embodiments, based on the recommendation relationship information, it may be determined which variables the information input into the impact factor analysis model needs to control, thereby grouping historical object resource data.

In some embodiments, the grouped historical object resource data is input into an influence factor analysis model, and based on the influence factor analysis model, data influence analysis is performed on to-be-verified influence data corresponding to the historical object resource data, so as to obtain an influence factor analysis result, wherein the influence factor analysis result represents influence distribution information of the historical object resource data of different groups on target recommendation index data, namely influence degree of the to-be-verified influence data on the target recommendation index data.

In some embodiments, referring to fig. 5, the influence factor analysis model includes an interference determination module and an influence analysis module, the history object resource data and the recommendation relationship information are input into the influence factor analysis model to perform data influence analysis, and the obtaining the influence factor analysis result includes:

s510, inputting the recommendation relation information into an interference determination module, and determining an interference node corresponding to an associated node of the target node based on path information between each node in the recommendation relation information and the target node corresponding to the target recommendation index data;

s520, inputting the related nodes, the interference nodes and the historical object resource data into an influence analysis module, and carrying out data influence analysis on the historical object resource data to obtain an influence factor analysis result.

In some embodiments, based on the recommendation relationship information, the associated node of the target node corresponding to the target recommendation index data, that is, the object associated data or the resource associated data having a direct influence on the target recommendation index data, may be determined. When determining the interference node corresponding to the associated node, the interference node corresponding to the associated node can be determined based on the path information between each node in the recommended relation information and the target node corresponding to the target recommended index data, and if the path information between a certain node and the target node is overlapped with the path information between the associated node and the target node or the path information between the associated node and the target node is overlapped, the interference node corresponding to the associated node can be determined, namely the node which affects the associated node when determining the target recommended index data.

For example, the node a may reach the node C through the node B, the node B may directly reach the node C, and the path from the node B to the node C is a combined path, where the node B is an associated node, and the node a is an interference node when the node C is a target node. For example, the node D may reach the node F through the node E, and the node D may also directly reach the node F, where the node D is a coincident node of two pieces of path information, and if the node D is an associated node and the node F is a target node, the node E is an interference node.

When the historical object resource data is classified, object association data or resource association data corresponding to the interference node is required to be processed, for example, the object association data or resource association data corresponding to the interference node is controlled to be unchanged, or interference is removed in other modes, and then the historical object resource data is classified based on the object association data or resource association data corresponding to the association node, so that the influence of the interference node on the association node is avoided.

In some embodiments, as shown in the causal relationship diagram in fig. 4, there is advertisement video duration distribution on a path from video content to advertisement-segment-uniform-broadcasting rate, so that the video content is an interference node of an association node corresponding to the advertisement video duration distribution, when determining the influence of the advertisement video duration distribution on the advertisement-segment-uniform-broadcasting rate, the video content needs to be controlled to be unchanged, for example, the historical object resource data is classified based on the video content, so as to obtain multiple categories such as games, living categories and the like, and the data related to the advertisement video duration distribution is acquired in each category as the historical object resource data.

In addition, the advertisement video duration distribution is located on a path of the user distribution returning to the advertisement uniform distribution rate, the advertisement video duration distribution is an interference node of an associated node corresponding to the user distribution, but because the user distribution affects the advertisement video duration distribution, the advertisement video duration distribution affects the advertisement uniform distribution rate, when the influence of the user distribution on the advertisement uniform distribution rate is determined, although the advertisement video duration distribution is the interference node, the advertisement video duration distribution cannot be controlled to be unchanged, otherwise, the influence of the user distribution on the advertisement uniform distribution rate is calculated inaccurately.

In some embodiments, the historical object resource data may include at least two sets of data, the historical object resource data may be a binary data set of experimental set data and control set data, the single-dimensional influence factor may be verified based on the binary data set, the target control resource data may be a multiple data set of multiple experimental set data and control set data, or a multiple data set corresponding to continuous user grouping, and the multi-dimensional influence factor may be verified based on the multiple data set.

In some embodiments, the influence factor analysis model may include a plurality of models such as causal forest, X-learner, S-learner, ATE-IPTW, etc., more multidimensional and informative features may be introduced into the model, and iterative optimization of the influence factor analysis model may be continued.

When the X-learner model is adopted for data influence analysis, in an influence analysis module, based on a base learner, function fitting is carried out on the comparison group data in the historical object resource data, and a function corresponding to the comparison group data and a corresponding function result are obtained, wherein the function result is the increment of target recommendation index data. Based on the basic learner, performing function fitting on the experimental group data in the historical object resource data to obtain a function corresponding to the experimental group data and a corresponding function result, wherein the function result is the increment of the target recommendation index data. And calculating first estimation information corresponding to the experimental group data based on a function corresponding to the control group data, and subtracting the first estimation information from the function result of the experimental group data to obtain a first output result. And calculating second estimation information corresponding to the comparison group data based on the function corresponding to the experimental group data, and subtracting the function result of the comparison group data from the second estimation information to obtain a second output result. The first output result and the second output result are weighted to obtain a result of evaluating a conditional processing effect (Conditional Average Treatment Effect, CATE) function, and an image corresponding to the machine learning interpretable tool (Shapley Additive explanation, shape) can be generated. Therefore, the X-learner model can utilize the thought that the observed sample result is used for estimating the unobserved sample result, approximate the increment of the target recommended index data, and simultaneously perform tendency weight adjustment on the first output result and the second output result so as to achieve the purpose of optimizing the approximate result.

In some embodiments, when the causal classification model is used for data impact analysis, decision trees may be used for grouping in the impact analysis module, and then, for each leaf inside, subtracting the average value of the experimental group data from the average value of the control group data, so as to obtain the CATE result.

In some embodiments, when the ATE-IPTW is used in the impact analysis module to perform data impact analysis, in the impact analysis module, a probability of accepting a process may be assigned to each historical object resource data, and then the result corresponding to each historical object resource data may be weighted according to its inverse probability, i.e., the actual processed historical object resource data is predicted to be largely unprocessed, and is given a greater weight than the predicted historical object resource data that is likely to be processed.

In some embodiments, in the case that the historical object resource data is an advertisement video, the following table shows the analysis results of the influence factors of the users with different age groups on the average advertising rate when the duration of the advertisement video is 0-20 s.

It can be found from the table that when the advertisement video duration is 0-20s, the total advertisement consumption rate is improved, wherein the improvement degree of the total advertisement rate of female users is larger than that of male users, and the improvement degree of the total advertisement rate is highest for females in the age range of 36-45 years.

Referring to fig. 6, fig. 6 shows a shape diagram corresponding to the analysis result of the influencing factor when the duration of the advertisement video is 0-20s in the case that the historical object resource data is the advertisement video. Based on the analysis results of the influence factors shown in fig. 6, it can be known that the improvement of the duty ratio of advertisement videos of living beings, such as cat-related videos shown in fig. 6, for example, has an improvement effect on the average completion rate of advertisement consumption of users, while the improvement of the duty ratio of advertisement videos of 0-20s has a reduction effect on the average completion rate of advertisement consumption of users. Female users show great interest preference on advertisement videos of 0-20s, and for users older than 55 years old, the advertisement videos of 0-20s account for the improvement of the ratio, and the improvement of the rate of the advertisement consumption of the users is achieved. Therefore, based on the analysis result of the influence factors, the business of the resource recommendation processing can be adjusted.

The method comprises the steps of determining interference nodes based on recommendation relation information, then carrying out data influence analysis on historical object resource data through an influence factor analysis model to obtain an influence factor analysis result, and verifying whether causality exists among related data by eliminating interference factors in the environment and carrying out causal inference on correlation between the influence data to be verified and target recommendation index data, so that accuracy and effectiveness of data influence analysis can be improved.

In some embodiments, referring to fig. 7, the input of the historical object resource data into the influence factor analysis model for data influence analysis, the obtaining of the influence factor analysis result includes:

s710, determining at least one target influence factor analysis model from a plurality of preset influence factor analysis models based on the data type in the historical object resource data and the preset data influence analysis scene;

s720, inputting historical object resource data into at least one target influence factor analysis model for data influence analysis, and obtaining a model analysis result corresponding to each target influence factor analysis model;

s730, carrying out data fusion processing on each model analysis result, and determining an influence factor analysis result.

In some embodiments, the pre-set impact factor analysis model may include a plurality of models including causal forest, X-learner, ATE-IPTG, and the like. The causal forest is a non-parametric method, and the effect is not deteriorated due to the increase of the dimension of the variable, so that when the historical object resource data is a multi-dimensional data set, a causal forest model can be adopted for data influence analysis, and the gradual distribution and the construction confidence interval of the historical object resource data can be obtained through the causal forest model.

The X-learner model is also effective when the experimental group data is much larger than the control group data, and takes advantage of the functional structural characteristics of CATE. The X-learner model is capable of rapidly producing images corresponding to the CATE results and the machine learning interpretable tool (ShapleyAdditive explanation, shape). ATE-IPTW may aggregate homogeneous objects in the event of object association data loss.

Thus, based on the data types in the historical object resource data and the preset data impact analysis scenario, at least one target impact analysis model may be determined from the plurality of impact analysis models. For example, in the case where the data type in the historical object resource data does not include the type of the object association data, a model corresponding to the ATE-IPTW matching method may be used, and in the case where the data impact analysis scene needs to obtain the CATE result and the shape map, an X-learner model may be used. And aggregating model analysis results corresponding to different influence factor analysis models by combining a plurality of influence factor analysis models to obtain the influence factor analysis results.

In some embodiments, multiple influence factor analysis models may be integrated in a causal inference tool box, which includes three models, namely, data input, model training and result output, and can determine an applicable influence factor analysis model for causal inference processing according to input historical object resource data to obtain an influence factor analysis result.

In some embodiments, the target influence factor analysis model includes a target interference determination module and a target influence analysis module, the recommendation relationship information is input into the target interference determination module, and interference nodes corresponding to associated nodes of the target nodes can be determined based on path information between each node in the recommendation relationship information and the target node corresponding to the target recommendation index data. And inputting the associated node, the interference node and the historical object resource data into a target influence factor analysis module, and carrying out data influence analysis on the historical object resource data to obtain a model analysis result corresponding to the target influence factor analysis model. The method for obtaining the analysis result of the influence factor can be referred to in the embodiment corresponding to the method for inputting the historical object resource data and the recommendation relationship information into the influence factor analysis model to perform the data influence analysis.

By aggregating the current commonly used influence factor analysis models, the corresponding influence factor analysis models can be obtained for analysis based on the historical object resource data of different structures, so that the use efficiency of the influence factor analysis models can be improved, and the influence factor analysis models are easier to apply to data influence analysis.

In some embodiments, referring to fig. 8, after determining the historical object resource data associated with the target recommendation index data from the historical object resource data based on the recommendation relationship information, the method further includes:

s810, performing at least one feature construction process on the historical object resource data to obtain at least one target feature data;

inputting the historical object resource data into an influence factor analysis model for data influence analysis, wherein the obtaining of the influence factor analysis result comprises the following steps:

s820, inputting at least one piece of target characteristic data into an influence factor analysis model to perform data influence analysis, and obtaining an influence factor analysis result.

In some embodiments, at least one target feature data, which is a feature useful for model prediction, may be obtained by performing at least one feature construction process on the historical object resource data based on the target and business logic that the influence factor analysis model needs to learn. Referring to fig. 9, as shown in fig. 9, when the feature construction is performed, feature construction processing is performed on the historical object resource data through an aggregation feature construction, a simple conversion feature construction, a cartesian product feature construction, a genetic programming feature construction, a gradient lifting decision tree (Gradient Boosting Decision Tree, GBDT) feature construction, a clustering feature construction, a temporal feature construction, a time sequence feature construction, a spatial feature construction, a text feature construction, an automation feature construction, and the like, so that new features useful for model prediction can be created, that is, target feature data can be obtained.

And inputting the target characteristic data obtained after the characteristic construction treatment into an influence factor analysis model for data influence analysis to obtain an influence factor analysis result.

After the characteristic construction processing is carried out on the historical object resource data, the characteristic dimension and the information quantity can be increased, and model prediction is carried out on the influence factor analysis model, so that the effectiveness of data influence analysis can be improved.

S240, executing resource recommendation processing based on the influence factor analysis result.

In some embodiments, based on the result of the influence factor analysis, the influence degree of each influence data to be verified on the target recommendation index data may be quantified, so that the business of the resource recommendation process is adjusted as a whole to execute the resource recommendation process.

In some embodiments, referring to fig. 10, performing the resource recommendation operation based on the impact factor analysis result further includes:

s1010, executing target resource recommendation operation corresponding to the influence factor analysis result on a sample object associated with the target object to obtain a sample recommendation result;

s1020, verifying an influence factor analysis result based on a sample recommendation result;

s1030, executing the step of executing the recommendation operation based on the influence factor analysis result under the condition that the influence factor analysis result passes the verification.

In some embodiments, the sample object associated with the target object is determined, and the sample object may be a preset number of target objects. And adjusting a resource recommendation processing service based on the influence factor analysis result, and executing target resource recommendation operation corresponding to the influence factor analysis result on the sample object to obtain a sample recommendation result.

When verifying the influence factor analysis result based on the sample recommendation result, comparing a first change degree of target recommendation index data corresponding to the influence factor analysis result with a second change degree of target recommendation index data corresponding to the sample recommendation result, and if the change trend of the first change degree is consistent with that of the second change degree, namely that the first change degree indicates that the target recommendation index data is lifted, the second change degree also indicates that the target recommendation index data is lifted, and at the moment, the forward influence of target resource recommendation operation corresponding to the influence factor analysis result on the target recommendation index data is explained, and the influence factor analysis result passes the verification. If the change trends of the first change degree and the second change degree are inconsistent, that is, if the first change degree indicates that the target recommendation index data is improved, the second change degree indicates that the target recommendation index data is reduced, and at the moment, it is indicated that the target resource recommendation operation corresponding to the influence factor analysis result has negative influence on the target recommendation index data, and then the influence factor analysis result fails to pass verification.

In the case where the influence factor analysis result passes the verification, the recommendation operation may be performed based on the influence factor analysis result in the actual resource recommendation service.

And verifying the analysis result of the influence factors, so that the resource recommendation service is adjusted under the condition that the statistics level and the practical application level obtain the consistent conclusion, and the credibility of the analysis result of the influence factors can be improved.

In some embodiments, referring to fig. 11, as shown in fig. 11, when applied in a scenario of advertisement video recommendation, the modules performing the resource recommendation method may include a dial test experiment module, a causal graph model, a data module, a model module, and a test module. In the dial testing experiment module, an experiment questionnaire can be designed, a dial testing experiment is carried out through a target user, analysis is carried out after experimental results are collected, influence data to be verified are determined, and therefore objects are clearly analyzed and enter the causal graph module. In the causal graph module, based on a first association relation between to-be-verified influence data and target recommendation index data and a second association relation between object association data and resource association data, nodes respectively corresponding to the object association data, the resource association data and the target recommendation index data are connected, so that recommendation relation information is constructed, and a causal graph is constructed. In the data module, a plurality of application data, namely historical object resource data, can be obtained, abnormal data in the application data are filtered and then subjected to feature construction processing, and a feature table is generated, so that the historical object resource data is obtained. The model module is integrated with a causal recommendation tool box and comprises a plurality of influence factor analysis models, and historical object resource data and recommendation relation information are input into the influence factor analysis models in the model module to obtain influence factor analysis results.

In the test module, a micro dial test experiment can be designed to verify whether target resource recommendation operation corresponding to the influence factor analysis result can promote target recommendation index data, and under the condition that verification is passed, A/B test can be performed based on the influence factor analysis result to verify again, the test result in the first case and the test result in the second case are compared, and whether the target resource recommendation operation corresponding to the influence factor analysis result is needed to be adopted is determined based on the test result to adjust resource recommendation service. For example, the target crowd is equally divided into a user corresponding to the case A and a user corresponding to the case B, the case A adopts target resource recommending operation corresponding to the influence factor analysis result, the conversion rate of the user is 30%, the case B does not adopt target resource recommending operation corresponding to the influence factor analysis result, the conversion rate of the user is 20%, the influence factor analysis result can promote target recommending index data, and it can be determined that the influence factor analysis result passes verification.

In some embodiments, the data modules, the model modules, and the data transmissions between the data modules and the model modules may be aggregated into a pipeline, and for tasks of a preset level, the data impact analysis may be automatically performed by only configuring the super parameters corresponding to each module.

The embodiment of the application provides a resource recommendation method, which comprises the following steps: and constructing and obtaining recommended relation information by taking target recommended index data, object associated data and resource associated data as nodes and taking the associated relation among the data corresponding to the nodes as edges, inputting historical object resource data and recommended relation information into an influence factor analysis model for data influence analysis, obtaining an influence factor analysis result, representing influence distribution information of the historical object resource data on the target recommended index data by the influence factor analysis result, and executing resource recommendation processing based on the influence factor analysis result. According to the method, the influence of each influence data on the target recommendation index data is quantized, so that recommendation analysis results can be returned to influence factor analysis results corresponding to different object groups, the accuracy and the effectiveness of influence factor analysis are improved, the influence factor analysis results can be applied to resource recommendation service scenes, and the rationality of resource recommendation services is improved.

The embodiment of the application also provides a resource recommendation device, please refer to fig. 12, which includes:

a historical data obtaining module 1210, configured to obtain historical object resource data corresponding to a target object, where the historical object resource data includes object association data of the target object and resource association data of a historical multimedia resource recommended to the target object;

The recommendation relationship information construction module 1220 is configured to construct recommendation relationship information by using the target recommendation index data, the object association data and the resource association data as nodes and the association relationship between the data corresponding to the nodes as edges;

the data impact analysis module 1230 is configured to input the historical object resource data and the recommendation relationship information into the impact factor analysis model for data impact analysis, so as to obtain an impact factor analysis result, where the impact factor analysis result characterizes impact distribution information of the historical object resource data on the target recommendation index data;

a resource recommendation module 1240 for performing a resource recommendation process based on the impact factor analysis result.

In some embodiments, the influence factor analysis model includes an interference determination module and an influence analysis module, the data influence analysis module including:

the first interference determining unit is used for inputting the recommended relation information into the interference determining module, and determining interference nodes corresponding to the associated nodes of the target nodes based on path information between each node in the recommended relation information and the target node corresponding to the target recommended index data;

the first influence analysis unit is used for inputting the related node, the interference node and the historical object resource data into the influence analysis module, and carrying out data influence analysis on the historical object resource data to obtain an influence factor analysis result.

In some embodiments, the data impact analysis module includes:

the target model determining unit is used for determining at least one target influence factor analysis model from a plurality of preset influence factor analysis models based on the data type in the historical object resource data and the preset data influence analysis scene;

the model analysis result acquisition unit is used for inputting the historical object resource data and the recommendation relation information into at least one target influence factor analysis model to perform data influence analysis, so as to obtain a model analysis result corresponding to each target influence factor analysis model;

and the data fusion unit is used for carrying out data fusion processing on each model analysis result and determining the influence factor analysis result.

In some embodiments, the target influence factor analysis model includes a target interference determination module and a target influence analysis module, and the model analysis result acquisition unit includes:

the second interference determining unit is used for inputting the recommended relation information into the target interference determining module and determining interference nodes corresponding to the associated nodes of the target nodes based on path information between each node in the recommended relation information and the target node corresponding to the target recommended index data;

The second influence analysis unit is used for inputting the related node, the interference node and the historical object resource data into the target influence factor analysis module, and carrying out data influence analysis on the historical object resource data to obtain a model analysis result corresponding to the target influence factor analysis model.

In some embodiments, the resource recommendation module includes:

the sample recommendation result acquisition unit is used for executing target resource recommendation operation corresponding to the influence factor analysis result on the sample object associated with the target object to obtain a sample recommendation result;

the verification unit is used for verifying the analysis result of the influence factors based on the sample recommendation result;

and the execution unit is used for executing the step of executing the recommendation operation based on the influence factor analysis result under the condition that the influence factor analysis result passes the verification.

In some embodiments, the recommendation relationship information construction module comprises:

the recommendation analysis unit is used for carrying out recommendation analysis on recommendation results corresponding to the historical object resource information to obtain to-be-verified influence data associated with target recommendation index data, wherein the to-be-verified influence data comprises object associated data which influence the target recommendation index data and resource associated data which influence the target recommendation index data;

The first association relation determining unit is used for determining a first association relation between the target recommendation index data and the influence data to be verified;

a second association relation determining unit for determining a second association relation between the object association data and the resource association data;

the recommendation relation information construction unit is used for constructing recommendation relation information by taking target recommendation index data, object association data and resource association data as nodes and taking a first association relation and a second association relation as sides.

In some embodiments, the apparatus further comprises:

the characteristic construction unit is used for carrying out characteristic construction processing on the historical object resource data at least once to obtain at least one target characteristic data;

the data impact analysis module comprises:

and the characteristic analysis unit is used for inputting at least one target characteristic data into the influence factor analysis model to perform data analysis so as to obtain an influence factor analysis result.

The device provided in the above embodiment can execute the method provided in any embodiment of the present application, and has the corresponding functional modules and beneficial effects of executing the method. Technical details not described in detail in the above embodiments may be referred to a resource recommendation method provided in any embodiment of the present application.

The present embodiment also provides a computer-readable storage medium, in which computer-executable instructions are stored, and the computer-executable instructions are loaded by a processor and execute a resource recommendation method according to the present embodiment.

The present embodiments also provide a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The computer instructions are read from the computer-readable storage medium by a processor of a computer device, and executed by the processor, cause the computer device to perform the methods provided in the various alternative implementations of resource recommendation described above.

The present embodiment also provides an electronic device, which includes a processor and a memory, where the memory stores a computer program, and the computer program is adapted to be loaded by the processor and execute a resource recommendation method according to the present embodiment.

The device may be a computer terminal, a mobile terminal or a server, and the device may also participate in forming an apparatus or a system provided by an embodiment of the present application. As shown in fig. 13, the server 13 may include one or more processors 1302 (shown in the figures as 1302a, 1302b, … …,1302 n) (the processor 1302 may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA or the like processing device), a memory 1304 for storing data, and a transmission device 1306 for communication functions. In addition, the method may further include: input/output interface (I/O interface), network interface. It will be appreciated by those of ordinary skill in the art that the configuration shown in fig. 13 is merely illustrative and is not intended to limit the configuration of the electronic device described above. For example, the server 13 may also include more or fewer components than shown in fig. 13, or have a different configuration than shown in fig. 13.

It should be noted that the one or more processors 1302 and/or other data processing circuits described above may be referred to herein generally as "data processing circuits. The data processing circuit may be embodied in whole or in part in software, hardware, firmware, or any other combination. Furthermore, the data processing circuitry may be a single stand-alone processing module, or incorporated in whole or in part into any of the other elements in the server 13.

The memory 1304 may be used to store software programs and modules of application software, and the processor 1302 may execute the software programs and modules stored in the memory 1304 by executing the program instructions/data storage device corresponding to the method according to the embodiments of the present application, so as to perform various functional applications and data processing, that is, implement a method for generating a time-series behavior capturing frame based on a self-attention network as described above. Memory 1304 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, memory 1304 may further include memory remotely located relative to processor 1302, which may be connected to server 13 via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The transmission means 1306 is used to receive or transmit data via a network. The specific example of the network described above may include a wireless network provided by a communication provider of the server 13. In one example, the transmission means 1306 comprises a network adapter (Network Interface Controller, NIC) which can be connected to other network devices via base stations so as to communicate with the internet. In one example, the transmission device 1306 may be a Radio Frequency (RF) module, which is configured to communicate with the internet wirelessly.

The present specification provides method operational steps as described in the examples or flowcharts, but may include more or fewer operational steps based on conventional or non-inventive labor. The steps and sequences recited in the embodiments are merely one manner of performing the sequence of steps and are not meant to be exclusive of the sequence of steps performed. In actual system or interrupt product execution, the methods illustrated in the embodiments or figures may be performed sequentially or in parallel (e.g., in the context of parallel processors or multi-threaded processing).

The structures shown in this embodiment are only partial structures related to the present application and do not constitute limitations of the apparatus to which the present application is applied, and a specific apparatus may include more or less components than those shown, or may combine some components, or may have different arrangements of components. It should be understood that the methods, apparatuses, etc. disclosed in the embodiments may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, and the division of the modules is merely a division of one logic function, and may be implemented in other manners, such as multiple units or components may be combined or integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or unit modules.

Based on such understanding, the technical solution of the present application may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.

Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative elements and steps are described above generally in terms of functionality in order to clearly illustrate the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.

The above embodiments are only for illustrating the technical solution of the present application, and not for limiting the same; although the application has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for recommending resources, the method comprising:

2. The resource recommendation method according to claim 1, wherein the influence factor analysis model includes an interference determination module and an influence analysis module, the inputting the historical object resource data and the recommendation relationship information into the influence factor analysis model for data influence analysis, and obtaining the influence factor analysis result includes:

inputting the recommendation relation information into the interference determination module, and determining an interference node corresponding to an associated node of the target node based on path information between each node in the recommendation relation information and the target node corresponding to the target recommendation index data;

inputting the associated node, the interference node and the historical object resource data into an influence analysis module, and carrying out data influence analysis on the historical object resource data to obtain an influence factor analysis result.

3. The resource recommendation method according to claim 1, wherein the inputting the historical object resource data and the recommendation relationship information into an influence factor analysis model for data influence analysis, and obtaining the influence factor analysis result comprises:

Determining at least one target influence factor analysis model from a plurality of preset influence factor analysis models based on the data type in the historical object resource data and a preset data influence analysis scene;

inputting the historical object resource data and the recommendation relation information into at least one target influence factor analysis model for data influence analysis to obtain a model analysis result corresponding to each target influence factor analysis model;

and carrying out data fusion processing on each model analysis result, and determining an influence factor analysis result.

4. The resource recommendation method according to claim 3, wherein the target influence factor analysis model includes a target interference determination module and a target influence analysis module, the inputting the historical object resource data and the recommendation relationship information into at least one target influence factor analysis model for data influence analysis, and obtaining a model analysis result corresponding to each target influence factor analysis model includes:

inputting the recommendation relation information into the target interference determining module, and determining an interference node corresponding to an associated node of the target node based on path information between each node in the recommendation relation information and the target node corresponding to the target recommendation index data;

Inputting the associated node, the interference node and the historical object resource data into a target influence factor analysis module, and carrying out data influence analysis on the historical object resource data to obtain a model analysis result corresponding to the target influence factor analysis model.

5. The resource recommendation method according to claim 1, wherein the performing a resource recommendation operation based on the influence factor analysis result includes:

executing target resource recommendation operation corresponding to the influence factor analysis result on the sample object associated with the target object to obtain a sample recommendation result;

verifying the influence factor analysis result based on the sample recommendation result;

and executing the step of executing the recommendation operation based on the influence factor analysis result in the case that the influence factor analysis result passes the verification.

6. The resource recommendation method according to claim 1, wherein the constructing recommendation relationship information by using the target recommendation index data, the object association data and the resource association data as nodes and the association relationship between the corresponding data of the nodes as edges comprises:

recommendation analysis is carried out on recommendation results corresponding to the historical object resource information, to-be-verified influence data associated with the target recommendation index data is obtained, and the to-be-verified influence data comprises object association data affecting the target recommendation index data and resource association data affecting the target recommendation index data;

Determining a first association relationship between the target recommendation index data and the influence data to be verified;

determining a second association relationship between the object association data and the resource association data;

and constructing and obtaining recommendation relation information by taking the target recommendation index data, the object association data and the resource association data as nodes and the first association relation and the second association relation as sides.

7. The resource recommendation method according to claim 1, wherein the step of inputting the history object resource data and the recommendation relationship information into an influence factor analysis model for data influence analysis, and before obtaining the influence factor analysis result, the method further comprises:

performing at least one feature construction treatment on the historical object resource data to obtain at least one target feature data;

inputting the historical object resource data and the recommendation relation information into an influence factor analysis model for data influence analysis, and obtaining an influence factor analysis result comprises the following steps:

and inputting the at least one target characteristic data and the recommended relation information into the influence factor analysis model to perform data influence analysis, so as to obtain the influence factor analysis result.

8. A resource recommendation device, the device comprising:

9. An electronic device comprising a processor and a memory, wherein the memory has stored therein at least one instruction or at least one program, the at least one instruction or the at least one program being loaded and executed by the processor to implement a resource recommendation method according to any of claims 1-7.

10. A computer readable storage medium, characterized in that the storage medium comprises a processor and a memory, in which at least one instruction or at least one program is stored, which is loaded and executed by the processor to implement a resource recommendation method according to any of the claims 1-7.

11. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the resource recommendation method of any of claims 1-7.