CN115062227B

CN115062227B - User behavior activity analysis method adopting artificial intelligence analysis and big data system

Info

Publication number: CN115062227B
Application number: CN202210785985.8A
Authority: CN
Inventors: 文建明
Original assignee: Promotion Technology Beijing Co ltd
Current assignee: Promotion Technology Beijing Co ltd
Priority date: 2022-07-06
Filing date: 2022-07-06
Publication date: 2023-01-10
Anticipated expiration: 2042-07-06
Also published as: CN115062227A

Abstract

The embodiment of the application provides a user behavior activity analysis method and a big data system adopting artificial intelligence analysis, wherein the user behavior activity analysis is carried out on candidate users on the basis of a reference behavior expression network, target field parameters of candidate activity expression fields in a candidate behavior feature cluster corresponding to the candidate users are obtained, content pushing is carried out on the candidate users on the basis of the target field parameters of the candidate activity expression fields, viewpoint feedback data of the candidate users for content pushing results are obtained, a pushing effect portrait of the content pushing results is generated, the content pushing is carried out on the candidate users by combining with the reference behavior expression network configured in advance, the pushing effect portrait of the content pushing results is evaluated by combining with the viewpoint feedback of the candidate users, and therefore an optimization reference basis of the reference behavior expression network is provided for product developers.

Description

User behavior activity analysis method adopting artificial intelligence analysis and big data system

Technical Field

The application relates to the technical field of artificial intelligence, in particular to a user behavior activity analysis method and a big data system adopting artificial intelligence analysis.

Background

In the field of personalized information push based on big data, the knowledge graph technology expresses the complicated knowledge field by data mining, reveals the dynamic development rule of the knowledge field, and provides practical and valuable reference for personalized content push. For example, by constructing a knowledge graph of user behavior activities of a large number of users, relevant content rules of the user behavior activities can be mined, and personalized content services are provided for the user behavior activities in a targeted manner.

Disclosure of Invention

In order to overcome at least the above disadvantages in the prior art, the present application aims to provide a user behavior activity analysis method and a big data system using artificial intelligence analysis.

In a first aspect, the present application provides a user behavior and activity analysis method using artificial intelligence analysis, which is applied to a big data system, where the big data system is in communication connection with a plurality of cloud service servers, and the method includes:

performing user behavior activity analysis on candidate users based on a reference behavior expression network to obtain target field parameters of each candidate activity expression field in a candidate behavior feature cluster corresponding to the candidate users, and performing content push to the candidate users based on the target field parameters of each candidate activity expression field, wherein the reference behavior expression network adopts a knowledge graph structure and comprises a plurality of behavior feature clusters and corresponding influence coefficients, each behavior feature cluster comprises user behavior activities, activity expression fields and field parameters, and the candidate behavior feature clusters comprise the same candidate user behavior activities and candidate activity expression fields and field parameters corresponding to different content source paths;

and obtaining viewpoint feedback data of the candidate user aiming at the content pushing result, and generating a pushing effect portrait of the content pushing result.

In a second aspect, an embodiment of the present application further provides a user behavior and activity analysis system using artificial intelligence analysis, where the user behavior and activity analysis system using artificial intelligence analysis includes a big data system and multiple cloud service servers in communication connection with the big data system;

the big data system is used for:

By adopting any one of the above embodiments, the user behavior activity analysis is performed on the candidate user based on the reference behavior expression network, the target field parameter of each candidate activity expression field in the candidate behavior feature cluster corresponding to the candidate user is obtained, the content is pushed to the candidate user based on the target field parameter of each candidate activity expression field, then the viewpoint feedback data of the candidate user for the content pushing result is obtained, and the pushing effect portrait of the content pushing result is generated, so that the content pushing is performed on the candidate user by combining with the pre-configured reference behavior expression network, and the pushing effect portrait of the content pushing result is evaluated by combining with the viewpoint feedback of the candidate user, so as to provide the optimized reference basis of the reference behavior expression network for the product developers.

Drawings

Fig. 1 is a schematic flow chart of a user behavior activity analysis method using artificial intelligence analysis according to an embodiment of the present invention.

Detailed Description

The following describes an architecture of a user behavior activity analysis system 10 using artificial intelligence analysis according to an embodiment of the present invention, where the user behavior activity analysis system 10 using artificial intelligence analysis may include a big data system 100 and a cloud service server 200 communicatively connected to the big data system 100. The big data system 100 and the cloud service server 200 in the user behavior activity analysis system 10 adopting artificial intelligence analysis may execute the user behavior activity analysis method adopting artificial intelligence analysis in cooperation with the following method embodiment, and the specific steps executed by the big data system 100 and the cloud service server 200 may refer to the detailed description of the following method embodiment.

The method for analyzing the user behavior activity by using the artificial intelligence analysis provided in this embodiment may be executed by the big data system 100, and the method for analyzing the user behavior activity by using the artificial intelligence analysis is described in detail below with reference to fig. 1.

Before further detailed description of the embodiments of the present application, terms and expressions referred to in the embodiments of the present application will be described, and the terms and expressions referred to in the embodiments of the present application will be used for the following explanation.

And (3) user behavior activity: in the actual online business service process, a user operates an event (e.g., a browsing event for browsing the associated elements of the page) according to a specific behavior of the event (e.g., a page update event).

E, electric commercial user behavior activity: refers to a certain e-commerce operating event.

Activity expression field: the expression field refers to a certain characteristic dimension of the user behavior activity, such as field dimensions of interest points, page attention points and the like of the e-commerce user behavior activity.

Field parameters: the specific field content of the user behavior activity object on the activity expression field is referred, for example, the interest point of the e-commerce user behavior activity A is an a-page object.

Behavior feature clustering: refers to a structure similar to (h, r, t), h representing a user behavioral activity, r representing an activity expression field, and t representing a field parameter. For example, (e-commerce user activity a, point of interest, a page object) represents a cluster of behavioral characteristics, and the point of interest representing e-commerce user activity a is an a page object.

Content source path: refer to the different content source paths of the field parameters.

Influence coefficient: the influence coefficient representing a specific influence weight, such as a behavior feature cluster, can be considered as the credibility of the behavior feature cluster.

Behavior expression network: also known as behavior expression knowledgemaps, have the framework of a knowledgemap.

Based on the above explanations of terms, the steps of the embodiments of the present application are described in detail below.

In the Process100, a candidate user is subjected to user behavior activity analysis based on a reference behavior expression network, a target field parameter of each candidate activity expression field in a candidate behavior feature cluster corresponding to the candidate user is obtained, and content push is performed on the candidate user based on the target field parameter of each candidate activity expression field.

The reference behavior expression network comprises a plurality of behavior feature clusters and corresponding influence coefficients, each behavior feature cluster comprises user behavior activities, activity expression fields and field parameters, and the candidate behavior feature clusters comprise the same candidate user behavior activities and candidate activity expression fields and field parameters corresponding to different content source paths.

In the Process200, viewpoint feedback data of the candidate user with respect to the content push result is acquired, and a push effect picture of the content push result is generated.

For example, positive opinion feedback data and negative opinion feedback data in the opinion feedback data may be extracted and then a push effect profile of the content push results may be generated in conjunction with the positive opinion feedback data and the negative opinion feedback data. For example, a ratio of the viewpoint number of each service field in the positive viewpoint feedback data to the viewpoint number corresponding to each service field in the negative viewpoint feedback data may be calculated, and then a target push effect portrait corresponding to the ratio may be determined based on a mapping relationship between each ratio range configured in advance and the push effect portrait. In addition, in other possible embodiments, other manners, such as an AI analysis manner, may also be adopted, but not limited thereto.

By adopting the technical scheme, the embodiment analyzes the user behavior activity of the candidate user based on the reference behavior expression network, obtains the target field parameter of each candidate activity expression field in the candidate behavior feature cluster corresponding to the candidate user, pushes the content to the candidate user based on the target field parameter of each candidate activity expression field, then obtains the viewpoint feedback data of the candidate user for the content pushing result, and generates the pushing effect portrait of the content pushing result, so that the content pushing is performed on the candidate user by combining with the pre-configured reference behavior expression network, and the pushing effect portrait of the content pushing result is evaluated by combining with the viewpoint feedback of the candidate user, so as to provide the optimized reference basis of the reference behavior expression network for product developers.

In the Process101, a reference behavior expression network is acquired.

For example, a reference behavior expression network is first obtained. The reference behavior expression network comprises a plurality of behavior feature clusters and corresponding influence coefficients, and each behavior feature cluster comprises user behavior activity, an activity expression field and a field parameter.

In some exemplary design considerations, process101 may be implemented by processes 1011-Process1015 in the embodiments described below. The processes 1011 to 1015 will be explained below.

In the Process1011, a plurality of template behavior feature clusters are acquired from the template user behavior data.

For example, a plurality of template behavior feature clusters may be determined from the template user behavior data.

In some exemplary design concepts, a plurality of template behavior feature clusters are obtained from template user behavior data, and the method is implemented based on the following steps: extracting concerned variables of the template user behavior data to obtain a plurality of concerned variable pairs, wherein each concerned variable pair comprises a user behavior activity concerned variable and a field parameter concerned variable; the following is performed for each variable pair of interest: carrying out relation classification processing on the concerned variable pairs to obtain the relation between the user behavior activity concerned variable and the field parameter concerned variable in the concerned variable pairs, and taking the relation as an activity expression field concerned variable; and combining the activity expression field concerned variables and the concerned variables with the covered user activity concerned variables and the covered field parameter concerned variables to obtain the template behavior characteristic cluster.

For example, when a plurality of template behavior feature clusters are obtained, a large amount of template user behavior data need to be obtained, and attention variable extraction is performed on the obtained template user behavior data to obtain a plurality of attention variable pairs, where each attention variable pair includes a user behavior activity attention variable and a field parameter attention variable.

For example, an attention variable pair (user behavior activity a and page object B) can be obtained by performing attention variable extraction on template user behavior data "user behavior activity a is a point of interest generated in page object B", where user behavior activity a represents a user behavior activity attention variable and page object B represents a field parameter attention variable.

After a plurality of concerned variable pairs are obtained, relation classification processing is carried out on each concerned variable pair, the relation between the user behavior activity concerned variable and the field parameter concerned variable in each concerned variable pair is obtained, and the obtained relation is used as the activity expression field concerned variable. And then configuring a template behavior characteristic cluster according to the concerned variable pair and the concerned variable of the activity expression field.

For example, a relation classification process is performed on the concerned variable pair (user behavior activity a, page object B), and the relation between the two concerned variables is obtained as a "page attention point", so that a template behavior feature cluster (user behavior activity a, page attention point, page object B) is configured according to the page attention point and the (user behavior activity a, page object B).

By adopting the technical scheme to obtain the template behavior feature clusters, a large number of template behavior feature clusters can be obtained, and an initial behavior expression network can be conveniently configured according to the large number of template behavior feature clusters.

In Process1012, the basic behavior expression network is configured by: generating network members corresponding to the plurality of template behavior characteristic clusters respectively; determining the connection relation of the first class members in the basic behavior expression network between every two template behavior feature clusters matched with the first matching state; and determining the connection relation of the second type of members in the basic behavior expression network between every two template behavior characteristic clusters matched with the second matching state.

For example, after obtaining a plurality of template behavior feature clusters, the basic behavior expression network is configured according to the template behavior feature clusters.

For example, taking the number of the template behavior feature clusters as 6 as an example, each template behavior feature cluster corresponds to one network member in the generated basic behavior expression network, so that 6 network members are generated.

And determining a first class member connection relation in the basic behavior expression network between every two template behavior feature clusters matched with a first matching state, wherein the first matching state is the content source path matching of the field parameters of the template activity expression fields covered by every two template behavior feature clusters. For example, assuming that the content source paths of the field parameters of the template activity expression fields covered by the template behavior feature clusters corresponding to the network members 1, 3 and 6 are matched, the template behavior feature clusters corresponding to each two network members of the network members 1, 3 and 6 are matched with a first matching state, so that a first class member connection relationship in the basic behavior expression network is generated among the network members 1, 3 and 6; the content source paths of the field parameters of the template activity expression fields covered by the template behavior feature clusters corresponding to the network members 2, 4 and 5 are matched, so that the template behavior feature clusters corresponding to every two network members in the network members 2, 4 and 5 are also matched with the first matching state, and therefore, the first-class member connection relationship in the basic behavior expression network is also generated among the network members 2, 4 and 5.

And determining the connection relation of the second class of members in the basic behavior expression network between every two template behavior feature clusters matched with the second matching state, wherein the second matching state is the matching of the template user behavior activity covered by every two template behavior feature clusters and the template activity expression field. The template user behavior activity and the template activity expression field covered by the template behavior characteristic clusters corresponding to the network member 1 and the network member 2 are matched, so that the template behavior characteristic cluster corresponding to the network member 1 and the template behavior characteristic cluster corresponding to the network member 2 are matched in a second matching state, and a second type of member connection relation in the basic behavior expression network is generated between the network member 1 and the network member 2; similarly, a second type of member connection relation in the basic behavior expression network is generated between the network member 3 and the network member 4; and generating a second class member connection relation in the basic behavior expression network between the network member 5 and the network member 6, thereby configuring the basic behavior expression network according to the 6 template behavior feature clusters.

In the Process1013, influence coefficients corresponding to the plurality of template behavior feature clusters are obtained, and influence coefficients corresponding to the first-class member connection relationship and the second-class member connection relationship are determined.

For example, after the basic behavior expression network is configured, the influence coefficient corresponding to each template behavior feature cluster, the influence coefficient corresponding to the first class member connection relationship, and the influence coefficient corresponding to the second class member connection relationship in the basic behavior expression network are obtained.

In some exemplary design ideas, before taking an influence coefficient of a hot spot behavior feature cluster in a plurality of template behavior feature clusters as a first influence coefficient, obtaining a hot spot user behavior activity by using a template user behavior activity in a basic behavior expression network, wherein the trigger frequency of the template user behavior activity is greater than a set frequency; acquiring one template behavior feature cluster from a plurality of template behavior feature clusters comprising each hotspot user behavior activity, and taking the acquired template behavior feature cluster as a hotspot behavior feature cluster corresponding to the hotspot user behavior activity; and acquiring a plurality of hotspot influence coefficients corresponding to each hotspot behavior feature cluster, and taking the mean influence coefficient of the hotspot influence coefficients as the influence coefficient corresponding to the hotspot behavior feature cluster.

For example, before determining the influence coefficient of each template behavior feature cluster in the basic behavior expression network, the template user behavior activity with the triggering frequency greater than the set frequency in the basic behavior expression network is determined, and it may be considered that the triggering frequency of the template user behavior activity in the basic behavior expression network is positively correlated with the reliability of the template user behavior activity, so that the template user behavior activity with the triggering frequency greater than the set frequency is taken as the hotspot user behavior activity.

For example, if there are 500 template user behavior activities in the basic behavior expression network, the trigger frequency of which is greater than the set frequency, then these 500 template user behavior activities are taken as the hotspot user behavior activities.

After determining the behavior activities of the hotspot users, each behavior activity of the hotspot users may correspond to a plurality of template behavior feature clusters, and in order to facilitate impact coefficient labeling, for example, one template behavior feature cluster is randomly obtained from the plurality of template behavior feature clusters including each behavior activity of the hotspot users, and the obtained template behavior feature cluster is used as the hotspot behavior feature cluster corresponding to the behavior activity of the hotspot user.

After determining the hotspot behavior feature cluster corresponding to each hotspot user behavior activity, acquiring a plurality of hotspot influence coefficients corresponding to each hotspot behavior feature cluster, and taking the mean influence coefficient of the plurality of hotspot influence coefficients as the influence coefficient corresponding to the hotspot behavior feature cluster. For example, 3 hotspot influence coefficients corresponding to each hotspot behavior feature cluster are obtained, and the mean influence coefficient of the 3 hotspot influence coefficients is used as the influence coefficient corresponding to the hotspot behavior feature cluster.

And the influence coefficient of the determined hotspot behavior characteristic cluster is in a reasonable range.

In some exemplary design ideas, obtaining influence coefficients corresponding to a plurality of template behavior feature clusters, and implementing the influence coefficients based on the following steps: taking an influence coefficient of a hot spot behavior feature cluster in the plurality of template behavior feature clusters as a first influence coefficient, wherein the influence coefficient of the hot spot behavior feature cluster is an influence coefficient of hot spot parameter statistics, the first influence coefficient is contained in a first influence coefficient interval, the first influence coefficient interval is a threshold closed interval formed by a first threshold and a second threshold, and the first threshold is smaller than the second threshold; and taking the influence coefficient of a non-hotspot behavior characteristic cluster in the plurality of template behavior characteristic clusters as a second influence coefficient, wherein the influence coefficient of the non-hotspot behavior characteristic cluster is the influence coefficient of hotspot parameter statistics, and the coefficient value of the second influence coefficient is the interval middle coefficient value of the first influence coefficient interval.

For example, after determining the hot spot behavior feature cluster, the influence coefficient of the hot spot behavior feature cluster is used as a first influence coefficient, and the first influence coefficient is an influence coefficient of hot spot parameter statistics. The coefficient value of the influence coefficient of the hotspot parameter statistics should be included in a uniformly set influence coefficient interval (e.g., a first influence coefficient interval), where the first influence coefficient interval is a threshold-value-enclosed interval composed of a first threshold value and a second threshold value, and the first threshold value is smaller than the second threshold value. For example, the first threshold value may be set to-1 and the second threshold value may be set to 1, so that the first influence coefficient is included in the first influence coefficient interval of [ -1,1].

The basic behavior expression network comprises a part of hotspot behavior characteristic clusters and a large number of non-hotspot behavior characteristic clusters, and the influence coefficient of each non-hotspot behavior characteristic cluster is used as a second influence coefficient. And the coefficient value of the second influence coefficient is the middle coefficient value of the interval of the first influence coefficient. The influence coefficients of the non-hotspot behavior feature clusters are set as the interval middle coefficient values of the first influence coefficient intervals, so that the influence coefficients of all the template behavior feature clusters counted by hotspot parameters are distributed more uniformly, the traversal optimization of the first loading network is carried out according to the influence coefficients of the uniformly distributed template behavior feature clusters, and the influence coefficients after the traversal optimization of the template behavior feature clusters are more reasonable. Continuing with the above example, the coefficient value of the second influence coefficient may be 0 on the condition that the first influence coefficient included in the first influence coefficient interval is [ -1,1].

By respectively taking the hot spot behavior characteristic cluster and the non-hot spot behavior characteristic cluster as different influence coefficients, the credibility of the behavior activities of the template users corresponding to the hot spot behavior characteristic cluster and the non-hot spot behavior characteristic cluster is different, so that the influence coefficients can be correspondingly set based on the credibility of the template behavior characteristic cluster, and the reasonability of the influence coefficients corresponding to the template behavior characteristic clusters is improved.

In some exemplary design ideas, the influence coefficient of a hotspot behavior feature cluster in a plurality of template behavior feature clusters is used as a first influence coefficient, and the method is implemented based on the following steps: determining an influence coefficient of a template behavior characteristic cluster of a first target magnitude in a basic behavior expression network; determining the influence coefficients corresponding to the connection relations of the first class members and the second class members in the basic behavior expression network; the first class of member connection relation is the member connection relation between every two template behavior feature clusters which are matched with the first matching state; the second class of member connection relation is the member connection relation between every two template behavior characteristic clusters which are matched with the second matching state; loading an influence coefficient corresponding to the template behavior feature cluster of the first target magnitude, an influence coefficient corresponding to the connection relation of the first class of members and an influence coefficient corresponding to the connection relation of the second class of members in the basic behavior expression network to obtain a second loading network; according to the influence coefficient corresponding to the template behavior feature cluster of the first target magnitude in the second loading network, the influence coefficient corresponding to the first class member connection relation and the influence coefficient corresponding to the second class member connection relation, traversing optimization is performed on the influence coefficient of each template behavior feature cluster in the second loading network, and second traversing optimization information is output and comprises: influence coefficients of each template behavior characteristic cluster after traversal optimization; taking the template behavior feature cluster with the influence coefficient larger than the set influence coefficient after traversal optimization as a hot spot behavior feature cluster; and taking the influence coefficient of the hot spot behavior feature cluster as a first influence coefficient.

For example, in response to an influence coefficient labeling operation for the template behavior feature cluster of the first target magnitude, receiving a hotspot influence coefficient for each of the template behavior feature clusters of the first target magnitude; and then, respectively determining the influence coefficients corresponding to the first class member connection relation and the second class member connection relation in the basic behavior expression network. The first class member connection relation is a member connection relation between every two template behavior feature clusters in a first matching state, namely, the content source paths of the field parameters of the template activity expression fields covered by every two template behavior feature clusters connected by the first class member connection relation are matched; the second type member connection relation is a member connection relation between every two template behavior feature clusters matching the second matching state, that is, the template user behavior activity and the template activity expression field covered by every two template behavior feature clusters connected by the second type member connection relation are matched.

After determining the influence coefficient of the template behavior feature cluster of the first target magnitude, the influence coefficient corresponding to the first class member connection relation and the influence coefficient corresponding to the second class member connection relation, loading the influence coefficient corresponding to the template behavior feature cluster of the first target magnitude, the influence coefficient corresponding to the first class member connection relation and the influence coefficient corresponding to the second class member connection relation in the basic behavior expression network, and obtaining a second loading network. The network architecture of the second loading network is matched with the network architecture corresponding to the basic behavior expression network, and relative to the basic behavior expression network, the second loading network is loaded with the influence coefficients of the network members and the influence coefficients of the member connection relations.

After the second loading network is obtained, according to the influence coefficient corresponding to the template behavior feature cluster of the first target magnitude in the second loading network, the influence coefficient corresponding to the connection relation of the first class member and the influence coefficient corresponding to the connection relation of the second class member, traversal optimization is carried out on the influence coefficient of each template behavior feature cluster in the second loading network, and second traversal optimization information is output and comprises the influence coefficient of each template behavior feature cluster after traversal optimization; and then, taking the template behavior feature cluster with the influence coefficient larger than the set influence coefficient after traversing optimization as a hot spot behavior feature cluster, and taking the influence coefficient of the hot spot behavior feature cluster as a first influence coefficient.

For example, the influence coefficient after traversal optimization is higher than the corresponding reliability of the template behavior feature cluster with the set influence coefficient, so that the template behavior feature cluster is used as a hot spot behavior feature cluster, and the influence coefficient after traversal optimization of the template behavior feature cluster is used as a first influence coefficient of the hot spot behavior feature cluster. Therefore, the influence coefficient of the hot spot behavior characteristic cluster determined by the scheme is more reasonable.

In some exemplary designs, process1013 in the aforementioned embodiments may be implemented by Process10131-Process 10132. The Process10131-Process10132 will be explained below.

In the Process10131, the influence coefficient of the connection relationship between the first class members between every two template behavior feature clusters matching the first matching state is obtained as a third influence coefficient.

For example, every two template behavior feature clusters connected by the first class member connection relationship match the first matching state, that is, every two template behavior feature clusters connected by the first class member connection relationship and matched with the content source path of the field parameter of the covered template activity expression field can be regarded as having a weak positive correlation between every two template behavior feature clusters connected by the first class member connection relationship, so that the influence coefficient corresponding to the first class member connection relationship can be used as the third influence coefficient, the third influence coefficient is included in the second influence coefficient interval, the second influence coefficient interval is a threshold open interval formed by the second influence coefficient and the second threshold, and the second influence coefficient is smaller than the second threshold.

For example, under the condition that the coefficient value of the second influence coefficient is 0 and the coefficient value of the second threshold value is 1, the second influence coefficient section in which the third influence coefficient is included is (0, 1). For example, the third influencing factor may have a value slightly greater than 0, for example 0.01.

In the Process10132, according to a heuristic global optimization algorithm, an influence coefficient of the connection relationship of the second type members between every two template behavior feature clusters matching the second matching state is obtained.

For example, under the condition that every two template behavior feature clusters match the second matching state, that is, under the condition that the template user behavior activity covered by every two template behavior feature clusters and the template activity expression field match, the influence coefficient of the second-class member connection relationship connecting such two template behavior feature clusters is determined through a heuristic global optimization algorithm.

By correspondingly setting the influence coefficients corresponding to the first class member connection relation and the second class member connection relation, the influence coefficients of different member connection relations can be determined in a targeted manner, so that the determined influence coefficients of the member connection relations are more reasonable.

In some exemplary design concepts, the Process10132 in the foregoing embodiments may be implemented by the Process 101321-the Process 101322. The Process101321-Process101322 will be explained below.

In the Process101321, at every two template behavior feature clusters, the influence coefficient of the second-class member connection relationship between every two template behavior feature clusters is determined based on the first strategy.

For example, if the template activity expression fields covered by every two template behavior feature clusters connected by the second class member connection relationship are class type template activity expression fields, such as interest points, publishers and the like, the influence coefficient of the second class member connection relationship between every two template behavior feature clusters is determined based on the first strategy.

In some exemplary design ideas, the influence coefficient of the connection relationship of the second class member between every two template behavior feature clusters is determined based on a first strategy, and the method is implemented based on the following steps: determining a first relevance cost value between field parameters of two template activity expression fields covered by every two template behavior characteristic clusters and a hot spot maximum coverage rate between the field parameters of the two template activity expression fields; determining a product of the first relevance cost value and a third threshold value, and determining a ratio of the product to the maximum coverage rate of the hot spot, wherein the third threshold value is greater than a second threshold value; subtracting the difference generated by the ratio from 1 to obtain a second relevance cost value between every two template behavior feature clusters; and obtaining the influence coefficient of the second-class member connection relation between every two template behavior feature clusters according to the second relevance cost value between every two template behavior feature clusters.

For example, when determining the influence coefficient of the second class member connection relation based on the first policy, first determining a first relevance cost value between two field parameters covered by every two template behavior feature clusters of the second class member connection relation. In the embodiment of the present application, the first relevance cost value represents the minimum number of times of conversion from the field parameter 1 of one template behavior feature cluster (e.g., template behavior feature cluster 1) to the field parameter 2 of another template behavior feature cluster (e.g., template behavior feature cluster 2).

After a first relevance cost value between two field parameters covered by every two template behavior feature clusters connected by the second type of member connection relation is determined, the maximum hot spot coverage rate between the two field parameters covered by every two template behavior feature clusters connected by the second type of member connection relation is obtained.

For example, after determining a first relevance cost value and a maximum hot spot coverage rate between two field parameters covered by every two template behavior feature clusters connected by the second class member connection relationship, calculating a second relevance cost value between the two template behavior feature clusters by the following formula:

l =1- [ first relevance cost value third threshold value/LX (field parameter 1, field parameter 2) ]

Wherein L represents a second relevance cost value between every two template behavior feature clusters, and a third threshold value is greater than the second threshold value, for example, under the condition that the coefficient value of the second threshold value is 1, the coefficient value of the third threshold value may be 2; LX (field parameter 1, field parameter 2) represents the maximum coverage of hot spots between two field parameters covered by every two behavior feature clusters of the template.

Continuing with the above example, under the condition that the first correlation cost value between the field parameter 1 covered by the template behavior feature cluster 1 and the field parameter 2 covered by the template behavior feature cluster 2 is 2, and the maximum coverage rate of the hot spot is 3, the second correlation cost value between the two template behavior feature clusters is 1- (2 × 2/3), that is, -0.33, calculated by the above formula.

And after determining a second relevance cost value between every two template behavior feature clusters, taking the determined second relevance cost value as an influence coefficient of a second class member connection relation connecting the two template behavior feature clusters.

By taking the second relevance cost value between the two template behavior characteristic clusters as the influence coefficient of the second class member connection relationship, the second relevance cost value can directly express the proximity degree between the field parameters covered by the two template behavior characteristic clusters connected by the second class member connection relationship, so that the influence coefficient of the second class member connection relationship determined by adopting the scheme is more matched with the field parameters covered by the template behavior characteristic clusters.

In the Process101322, under the condition that the template activity expression fields covered by every two template behavior feature clusters are numerical template activity expression fields, determining an influence coefficient of the connection relationship of the second class member between every two template behavior feature clusters based on a second strategy.

For example, if the template activity expression fields covered by every two template behavior feature clusters are numerical template activity expression fields, such as page concerns, the influence coefficient of the second-class member connection relationship between every two template behavior feature clusters is determined based on the second strategy.

The influence coefficients of the connection relation of the second type members can be determined in a more targeted manner by adopting different modes to respectively determine the influence coefficients of the connection relation of the second type members based on the difference of the types of the template activity expression fields covered by every two template behavior characteristic clusters connected by the second type members, so that the determined influence coefficients of the connection relation of the second type members are more accurate.

In some exemplary design ideas, determining an influence coefficient of a second class member connection relationship between every two template behavior feature clusters based on a second strategy is implemented based on the following steps: under the condition that field parameters of two template activity expression fields covered by every two template behavior feature clusters are matched, obtaining a fourth influence coefficient which is the influence coefficient of the connection relation of the second type of members between every two template behavior feature clusters, wherein the coefficient value of the fourth influence coefficient is a second threshold value; and under the condition that the field parameters of the two template activity expression fields covered by every two template behavior characteristic clusters are not matched, obtaining a fifth influence coefficient which is the influence coefficient of the second class member connection relation between every two template behavior characteristic clusters, wherein the coefficient value of the fifth influence coefficient is a first threshold value.

For example, when determining the influence coefficient of the second class member connection relationship based on the second policy, the template activity expression field covered by every two template behavior feature clusters connected by the second class member connection relationship is a numerical template activity expression field, so that under the condition that the parameters of the two fields covered by the two template behavior feature clusters are matched, the influence coefficient of the second class member connection relationship between the two template behavior feature clusters is obtained as a fourth influence coefficient, and the coefficient value of the fourth influence coefficient is a second threshold value. For example, under the condition that the second threshold coefficient value is 1, the fourth influence coefficient has a coefficient value of 1.

And under the condition that the two field parameters covered by the two template behavior characteristic clusters are not matched, obtaining that the influence coefficient of the second class member connection relation between the two template behavior characteristic clusters is a fifth influence coefficient, and the coefficient value of the fifth influence coefficient is a first threshold value. For example, the value of the fifth influencing coefficient is-1 on the condition that the first threshold coefficient value is-1.

When determining the influence coefficient of the second class member connection relationship based on the second policy, the template user behavior activity and the template activity expression field covered by every two template behavior feature clusters connected by the second class member connection relationship are matched, in this case, if the field parameters covered by the two template behavior feature clusters are also matched, the two template behavior feature clusters are actually two matched template behavior feature clusters, so the influence coefficient of the second class member connection relationship connecting the two matched template behavior feature clusters is set to a larger value (i.e., a second threshold value); and if the field parameters covered by the two template behavior feature clusters do not match, so that the influence coefficient of the connection relation of the second class member between the two template behavior feature clusters with unmatched connections is set to be a smaller value (namely, the first threshold value is smaller than the second threshold value). Therefore, the influence coefficient of the connection relation of the second class members determined by the scheme is more reasonable.

In the Process1014, a plurality of influence coefficients corresponding to the template behavior feature clusters, the first class member connection relations and the second class member connection relations are loaded in the basic behavior expression network, and a first loading network is determined.

For example, after determining the influence coefficients corresponding to the plurality of template behavior feature clusters, the influence coefficients corresponding to the first class member connection relationships, and the influence coefficients corresponding to the second class member connection relationships in the basic behavior expression network, the influence coefficients corresponding to the plurality of template behavior feature clusters, the influence coefficients corresponding to the first class member connection relationships, and the influence coefficients corresponding to the second class member connection relationships are loaded in the basic behavior expression network, so as to obtain a first loading network.

The network architecture of the first loading network is matched with the network architecture corresponding to the basic behavior expression network, and relative to the basic behavior expression network, the first loading network is loaded with the influence coefficient of the network member and the influence coefficient of the member connection relation.

In the Process1015, traversal optimization is performed on the first loading network, and a reference behavior expression network is generated according to the generated first traversal optimization information.

For example, after the first loading network is obtained, traversal optimization is performed on the first loading network, first traversal optimization information is output, and then the reference behavior expression network is generated according to the first traversal optimization information. Wherein the first traversal optimization information includes: and (5) the influence coefficient of each template behavior characteristic cluster after traversal optimization.

When the influence coefficients of all the network members in the first loading network are converged, the converged traversing optimized influence coefficients of all the network members (namely all the template behavior feature clusters) are obtained, and a reference behavior expression network is configured according to the traversing optimized influence coefficients of all the template behavior feature clusters.

In some exemplary design considerations, traversal optimization of the first loading network in the Process1015 in the foregoing embodiment may be implemented by the Process10151 and the Process10152, and the Process10151 and the Process10152 are described below.

In the Process10151, a plurality of connected template behavior feature clusters corresponding to the template behavior feature clusters in the first loading network are obtained, and the influence coefficient product corresponding to each connected template behavior feature cluster is determined.

For example, in each traversal optimization process, a plurality of connected template behavior feature clusters corresponding to the template behavior feature clusters in the first loading network are determined, and an influence coefficient product corresponding to each connected template behavior feature cluster is determined. The connected template behavior feature cluster is directly connected with the template behavior feature cluster in the first loading network through member connection relation.

For example, the product of the influence coefficients corresponding to each connected template behavior feature cluster is the product of the influence coefficient of the connected template behavior feature cluster and the influence coefficient of the member connection relationship between the connected template behavior feature cluster and the template behavior feature cluster, and under the condition that the connected template behavior feature cluster and the template behavior feature cluster are matched in the first matching state, that is, under the condition that the connected template behavior feature cluster and the content source path of the field parameter of the template activity expression field covered by the template behavior feature cluster are matched, the influence coefficient of the member connection relationship is the influence coefficient corresponding to the member connection relationship of the first class; under the condition that the connected template behavior feature cluster and the template behavior feature cluster are matched in a second matching state, namely under the condition that the template user behavior activity and template activity expression fields covered by the connected template behavior feature cluster and the template behavior feature cluster are matched, the influence coefficient of the member connection relation is the influence coefficient corresponding to the second type of member connection relation.

In the Process10152, the influence coefficients optimized for the traversal of the template behavior feature cluster are obtained by adding the products of the influence coefficients corresponding to the plurality of connected template behavior feature clusters.

For example, after determining the influence coefficient products of each connected template behavior feature cluster corresponding to the template behavior feature cluster, summing the influence coefficient products of the connected template behavior feature clusters, and taking the summed result as the influence coefficient of the template behavior feature cluster after traversal optimization.

For example, assuming that the connected network members of the network member 6 include the network member 1, the network member 3, and the network member 5, in one traversal optimization process, the influence coefficient after traversal optimization of the network member 6 = the influence coefficient of the connected network member 1+ the influence coefficient of the member connection relationship between the network members 1 to 6 + the influence coefficient of the connected network member 3 + the influence coefficient of the member connection relationship between the network members 3 to 6 + the influence coefficient of the connected network member 5+ the influence coefficient of the member connection relationship between the network members 5 to 6, that is, 0.01+0 (-1), so that after the traversal optimization process is completed, the influence coefficient after traversal optimization of the network member 6 is 0.

The influence coefficient of the template behavior characteristic cluster is updated according to the influence coefficient of the connected network members in each traversal optimization process, so that the influence coefficient of the traversal optimized template behavior characteristic cluster is more accurate.

In some exemplary design concepts, the reference behavior expression network is determined according to the generated first traversal optimization information, and the determination is implemented based on the following steps: in a plurality of template behavior feature clusters comprising the same template user behavior activity and template activity expression fields, traversing the optimized template behavior feature cluster with the largest influence coefficient to obtain a first target behavior feature cluster corresponding to the template user behavior activity; and determining a reference behavior expression network according to the first target behavior characteristic clusters corresponding to the behavior activities of the template users covered by the first loading network and the influence coefficients corresponding to the first target behavior characteristic clusters.

When the Process1011 obtains the template behavior feature cluster from the template user behavior data, a plurality of conflicting template behavior feature clusters may be obtained, such as (user behavior activity AA, interest point, page object a) and (user behavior activity AA, interest point, page object b). And the plurality of conflict template behavior feature clusters are all used for configuring a basic behavior expression network, and the first loading network is generated according to the basic behavior expression network, so that the first loading network comprises the plurality of conflict template behavior feature clusters, and after the first loading network is subjected to traversal optimization, an accurate template behavior feature cluster can be determined from the plurality of conflict template behavior feature clusters.

For example, an accurate template behavior feature cluster may be determined from a plurality of conflicting template behavior feature clusters based on the following steps: and aiming at a plurality of template behavior characteristic clusters comprising the same template user behavior activity and template activity expression fields in the first loading network, traversing the optimized template behavior characteristic cluster with the largest influence coefficient to obtain a first target behavior characteristic cluster corresponding to the template user behavior activity. Because the influence coefficient of the first target behavior feature cluster is the largest, the first target behavior feature cluster is used as a behavior feature cluster accurately corresponding to the behavior activity of the template user.

After the first target behavior feature clusters corresponding to the behavior activities of the template users in the first loading network are determined by the same scheme, the reference behavior expression network is determined according to the first target behavior feature clusters and the influence coefficients corresponding to the first target behavior feature clusters.

For example, from all the template behavior feature clusters including the user behavior activity "user behavior activity AA" and the activity expression field "interest point", the template behavior feature cluster with the largest influence coefficient after traversal optimization is selected as the first target behavior feature cluster corresponding to the user behavior activity "user behavior activity AA", that is, the behavior feature cluster accurately corresponding to the user behavior activity "user behavior activity AA". For example, 2 template behavior feature clusters in the first loading network include a user behavior activity "user behavior activity AA" and an activity expression field "interest point", where the 2 template behavior feature clusters are respectively (user behavior activity AA, interest point, page object a) and (user behavior activity AA, interest point, page object b), and since the influence coefficient of the template behavior feature cluster (user behavior activity AA, interest point, page object a) after traversal optimization is-0.3, and the influence coefficient of the template behavior feature cluster (user behavior activity AA, interest point, page object b) after traversal optimization is 0.8, the most optimized influence coefficient (user behavior activity AA, interest point, page object b) is taken as a first target behavior feature cluster corresponding to the user behavior activity "user behavior activity AA".

For example, the reference behavior expression network is configured by: generating network members corresponding to the first target behavior feature clusters respectively; determining a first class member connection relation between every two first target behavior characteristic clusters matched with the content source path of the field parameter of the covered activity expression field; and determining the connection relation of the second class members between every two first target behavior characteristic clusters matched with the covered user behavior activities.

According to the influence coefficient after traversal optimization of the template behavior feature clusters, a first target behavior feature cluster is determined from a plurality of template behavior feature clusters comprising the same template user behavior activity and template activity expression fields, a reference behavior expression network is configured according to the plurality of first target behavior feature clusters, and each first target behavior feature cluster is an accurate template behavior feature cluster determined from a plurality of conflicting template behavior feature clusters, so that the accuracy of the reference behavior expression network determined by the scheme is higher.

In the Process102, the influence coefficients of the activity expression fields on the associated content source paths are obtained according to the influence coefficients of the behavior feature clusters.

For example, after the reference behavior expression network is obtained, the influence coefficients of each behavior feature cluster in the reference behavior expression network are obtained, and the influence coefficients of each activity expression field on the associated content source paths are obtained according to the influence coefficients of each behavior feature cluster. And the plurality of content source paths associated with the activity expression fields are content source paths of different field parameters of the activity expression fields.

In some exemplary design considerations, process102 in the foregoing embodiments may be implemented by Process1021-Process 1023. Next, processes 1021 to 1023 will be explained.

In Process1021, a first behavior feature cluster in the reference behavior expression network is obtained.

For example, when determining the influence coefficient of each activity expression field on the associated content source path, the activity expression field is walked, each content source path associated with the activity expression field currently walked to is walked, and the behavior feature cluster including the activity expression field currently walked to is taken as the first behavior feature cluster.

In Process1022, a second behavior feature cluster is obtained, and a second target magnitude of the second behavior feature cluster is determined.

For example, the second behavior feature cluster includes an activity expression field currently walked to, and the corresponding field parameter comes from the content source path currently walked to, in other words, the corresponding field parameter comes from the first behavior feature cluster of the content source path currently walked to, which is obtained as the second behavior feature cluster, and the number of the second behavior feature clusters in the basic behavior expression network is determined to be the second target magnitude.

In the Process1023, the ratio of the sum of the influence coefficients of the second behavior feature clusters to the second target magnitude is used to obtain the influence coefficient of the currently walked activity expression field on the currently walked content source path.

For example, after all the second behavior feature clusters are determined, the influence coefficients of all the second behavior feature clusters are added, and the ratio of the generated addition result to the second target magnitude corresponding to the second behavior feature cluster is used to obtain the influence coefficient of the currently wandering activity expression field on the currently wandering content source path.

Because the influence coefficient of the activity expression field currently walked on the content source path currently walked is the mean influence coefficient of the influence coefficients of a plurality of behavior feature clusters including the activity expression field currently walked and corresponding field parameters from the content source path currently walked, the influence coefficient of the activity expression field currently walked on the content source path currently walked determined by adopting the scheme is more accurate.

In the Process103, a content source path sequence of the plurality of content source paths associated with each active expression field is obtained according to the influence coefficient of each active expression field on the associated plurality of content source paths.

For example, after determining the influence coefficients of the activity expression fields on the associated content source paths, the content source path sequence of the content source paths associated with the activity expression fields is obtained according to the influence coefficients of the activity expression fields on the associated content source paths.

In some exemplary design considerations, process103 in the foregoing embodiments may be implemented by Process1031-Process 1032. Process1031 to Process1032 will be explained below.

In the Process1031, for a plurality of content source paths associated with the activity expression field, the influence coefficients of the activity expression field on each content source path are obtained.

For example, when determining a content source path sequence of a plurality of content source paths associated with one activity expression field (e.g., a first activity expression field), first, for the plurality of content source paths associated with the first activity expression field, an influence coefficient of the first activity expression field on each content source path is obtained.

In the Process1032, the influence coefficients of the active expression fields on the plurality of content source paths are sorted in descending order, and a content source path sequence of the active expression fields on the plurality of content source paths is output.

For example, after determining the influence coefficients of the first activity expression field on each content source path, sorting the influence coefficients of the first activity expression field on the plurality of content source paths in a descending order, so as to determine a content source path sequence of the first activity expression field on the plurality of content source paths according to the sorting result in the descending order.

For example, there are 2 behavior feature clusters in the reference behavior expression network, which include activity expression field 1 and whose field parameters of activity expression field 1 come from content source path a, content source path B and content source path C, these two behavior feature clusters are behavior feature cluster 1 and behavior feature cluster 2, respectively, the influence coefficient of activity expression field 1 in behavior feature cluster 1 on content source path a is 0.5, the influence coefficient on content source path B is 0.4, the influence coefficient on content source path C is 0.1, the influence coefficient of activity expression field 1 in behavior feature cluster 2 on content source path a is 0.4, the influence coefficient on content source path B is 0.4, and the influence coefficient on content source path C is 0.2, then based on that the influence coefficient of activity expression field 1 on content source path a can be found to be (0.5 + 0.4)/2, that is 0.45; the influence coefficient of activity expression field 1 on content source path B is (0.4 + 0.4)/2, i.e. 0.4; the influence coefficient of activity expression field 1 on content source path C is (0.1 + 0.2)/2, i.e. 0.15. Then sorting the influence coefficients of the active expression field 1 on the three content source paths in a descending order can obtain content source path a (0.45) > content source path B (0.4) > content source path C (0.15), so that the content source path sequence of the three content source paths associated with the active expression field 1 is content source path a > content source path B > content source path C.

The content source path importance of the activity expression fields on the content source paths is obtained by sorting the results according to the order from large to small of the influence coefficients of the activity expression fields on the content source paths, in other words, if the influence coefficient of the activity expression field on a certain content source path is larger, the content source path importance corresponding to the content source path is higher, and therefore, the scheme can be adopted to accurately determine the content source path sequence corresponding to each activity expression field.

In the Process104, a plurality of candidate behavior feature clusters are obtained.

For example, after determining a content source path sequence of a plurality of content source paths associated with each activity expression field in the reference behavior expression network, a plurality of candidate behavior feature clusters may be obtained, and a target field parameter corresponding to a candidate activity expression field in the candidate behavior feature clusters is obtained. The plurality of candidate behavior feature clusters comprise the same candidate user behavior activity and candidate activity expression fields and field parameters corresponding to different content source paths.

In the Process105, a content source path sequence of different content source paths corresponding to the candidate active expression fields is obtained from the content source path sequences of the multiple content source paths associated with each active expression field.

For example, after a plurality of candidate behavior feature clusters are obtained, based on candidate activity expression fields covered by the candidate behavior feature clusters, content source path sequences of different content source paths corresponding to the candidate activity expression fields are obtained from content source path sequences of a plurality of content source paths associated with each activity expression field.

In the Process106, the content source path with the maximum influence of each field parameter of the candidate activity expression field is obtained according to the content source path sequence of different content source paths corresponding to the candidate activity expression field, and the target field parameter serving as the candidate activity expression field is obtained based on the field parameter of the content source path with the maximum influence.

For example, after a content source path sequence of different content source paths corresponding to the candidate activity expression field is obtained, since a plurality of candidate behavior feature clusters exist, the candidate activity expression field corresponds to a plurality of field parameters, and each field parameter corresponds to a different content source path, therefore, based on the obtained content source path sequence, a content source path with the maximum influence among the plurality of content source paths corresponding to the plurality of field parameters of the candidate activity expression field is obtained, and based on the field parameter of the content source path with the maximum influence, a target field parameter which is the candidate activity expression field is obtained.

For example, assume that in the reference row expression network, the content source path sequence of the plurality of content source paths associated with activity expression field 1 is content source path a > content source path B > content source path C; two candidate behavior feature clusters are assumed to exist, namely (electric commercial user behavior activity 1, activity expression field 1 and field parameter 1) and (electric commercial user behavior activity 1, activity expression field 1 and field parameter 2), wherein the field parameter 1 comes from a content source path A, and the field parameter 2 comes from a content source path C; and acquiring a content source path sequence, a content source path A, a content source path B and a content source path C corresponding to the activity expression field 1, wherein based on the content source path sequence, as the importance ranking of the content source path A is higher than that of the content source path C, the field parameter 1 based on the content source path A is used as the target field parameter of the candidate activity expression field 1 in the candidate behavior characteristic cluster.

In some exemplary design considerations, process106 in the foregoing embodiments may also include Process107-Process111. Next, the processes 107 to 111 will be explained.

In the Process107, the candidate behavior feature cluster including the target field parameter is taken as the second target behavior feature cluster.

For example, after determining the target field parameters corresponding to the candidate activity expression fields in the candidate behavior feature cluster, the candidate behavior feature cluster including the target field parameters is used as the second target behavior feature cluster.

In the Process108, the second target behavior feature cluster is loaded to the reference behavior expression network as a new network member, and is connected with a third behavior feature cluster covered in the reference behavior expression network to generate a third behavior expression network.

For example, after the second target behavior feature cluster is obtained, the second target behavior feature cluster is loaded into the reference behavior expression network as a new network member, and the second target behavior feature cluster and the third behavior feature cluster are connected in the reference behavior expression network, so as to obtain a third behavior expression network. And matching the user behavior activities covered by the third behavior characteristic cluster with the user behavior activities covered by the second target behavior characteristic cluster, or matching the content source path of the field parameter of the covered activity expression field with the content source path of the field parameter of the activity expression field covered by the second target behavior characteristic cluster.

And adding a second target behavior feature cluster corresponding to the candidate behavior feature cluster into the reference behavior expression network to obtain a third behavior expression network, wherein the third behavior expression network can be expanded to obtain the third behavior expression network with richer feature quantity.

In the Process109, under the condition that data synchronization is not performed, the influence coefficients corresponding to the plurality of behavior feature clusters, the first class member connection relationships and the second class member connection relationships in the third behavior expression network are obtained.

For example, after the third behavior expression network is obtained, the influence coefficients corresponding to the multiple behavior feature clusters, the influence coefficients corresponding to the first class member connection relationships, and the influence coefficients corresponding to the second class member connection relationships in the third behavior expression network may be obtained without performing data synchronization. Wherein the condition that data synchronization is not performed is that a specific state of a content push process for the candidate behavior feature cluster is not triggered. Correspondingly, the condition of data synchronization is to start a specific state of a content push process for triggering response to the candidate behavior feature cluster.

For example, the influence coefficient of the behavior feature cluster originally belonging to the reference behavior expression network in the third behavior expression network is the influence coefficient of the corresponding behavior feature cluster covered by the reference behavior expression network, and the influence coefficient of the behavior feature cluster not belonging to the reference behavior expression network in the third behavior expression network can be determined by means of hot spot parameter statistics.

The third step is to express a determination manner of the influence coefficients of the connection relationship between the first class member and the second class member in the network, which is similar to the determination manner of the influence coefficients of the connection relationship between the first class member and the second class member in the Process1013, and is not described herein again.

The first-class member connection relation is a member connection relation between every two behavior feature clusters in a first matching state, namely, the content source path matching of field parameters of activity expression fields covered by every two behavior feature clusters connected by the first-class member connection relation is realized; the second type of member connection relationship is a member connection relationship between every two behavior feature clusters matching the third requirement, that is, the user behavior activity covered by every two behavior feature clusters connected by the second type of member connection relationship is matched.

In the Process110, a plurality of influence coefficients corresponding to the behavior feature clusters, the first class member connection relations and the second class member connection relations are loaded in the third behavior expression network, and a third loading network is determined.

For example, after obtaining the influence coefficients corresponding to the multiple behavior feature clusters, the influence coefficients corresponding to the first class member connection relationships, and the influence coefficients corresponding to the second class member connection relationships in the third behavior expression network, the influence coefficients corresponding to the multiple behavior feature clusters, the influence coefficients corresponding to the first class member connection relationships, and the influence coefficients corresponding to the second class member connection relationships are loaded in the third behavior expression network, so as to obtain a third loading network.

In the Process111, traversal optimization is performed on the third loading network, and an optimized reference behavior expression network is determined according to the generated third traversal optimization information.

For example, after the third loading network is obtained, traversal optimization is performed on the third loading network, and third traversal optimization information is output, where the third traversal optimization information includes: and (5) traversing and optimizing the influence coefficient of each behavior feature cluster. And then determining an optimized reference behavior expression network according to the third traversal optimization information.

The manner of performing traversal optimization on the third loading network is similar to the manner of performing traversal optimization on the first loading network in the Process1015, and is not described herein again.

Based on the above steps, the present embodiment may continuously iterate and loop to execute the processes 109 to 110, so as to loop to obtain the optimized reference behavior expression network.

In some embodiments, big data system 100 may include a processor 110, a machine-readable storage medium 120, a bus 130, and a communication unit 140.

The processor 110 may perform various suitable actions and processes in accordance with a program stored in the machine-readable storage medium 120, such as program instructions associated with the user behavioral activity analysis methods employing artificial intelligence analysis described in the foregoing embodiments. The processor 110, the machine-readable storage medium 120, and the communication unit 140 perform signal transmission through the bus 130.

In particular, the processes described in the above exemplary flow diagrams may be implemented as computer software programs, according to embodiments of the present invention. For example, embodiments of the invention include a computer program product comprising a computer program embodied on a computer-readable medium, the computer program comprising program code for performing the method illustrated by the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network via the communication unit 140, and when executed by the processor 110, performs the above-described functions defined in the methods of the embodiments of the present invention.

Yet another embodiment of the present invention further provides a computer-readable storage medium, in which computer-executable instructions are stored, and when executed by a processor, the computer-executable instructions are used to implement the method for analyzing user behavior activity by using artificial intelligence analysis according to any of the above embodiments.

Yet another embodiment of the present invention further provides a computer program product, which includes a computer program, and when being executed by a processor, the computer program implements the method for analyzing user behavior and activity by using artificial intelligence analysis according to any of the above embodiments.

It should be understood that, although the various operation steps are indicated by arrows in the flow chart of the embodiment of the present invention, the implementation order of the steps is not limited to the order indicated by the arrows. In some implementation scenarios of embodiments of the present invention, the implementation steps in the flowcharts may be performed in other sequences as desired, unless explicitly stated otherwise herein. In addition, some or all of the steps in each flowchart may include several sub-steps or several stages according to an actual implementation scenario. Some or all of these sub-steps or stages may be performed at the same time, and individual ones of these sub-steps or stages may also be performed at different times, respectively. In a scenario where the execution time is different, the execution sequence of the sub-steps or phases may be flexibly configured according to requirements, which is not limited in the embodiment of the present invention.

The foregoing is only an alternative embodiment of a part of implementation scenarios of the present invention, and it should be noted that those skilled in the art should also be able to protect the scope of the embodiments of the present invention based on other similar implementation means according to the technical idea of the present invention without departing from the technical idea of the present invention.

Claims

1. A user behavior and activity analysis method adopting artificial intelligence analysis is applied to a big data system, and the method comprises the following steps:

obtaining viewpoint feedback data of the candidate user aiming at a content pushing result, and generating a pushing effect portrait of the content pushing result;

the step of analyzing the user behavior activity of the candidate user based on the reference behavior expression network to obtain the target field parameter of each candidate activity expression field in the candidate behavior feature cluster corresponding to the candidate user, and pushing the content to the candidate user based on the target field parameter of each candidate activity expression field comprises the following steps:

acquiring a reference behavior expression network;

obtaining the influence coefficient of each activity expression field on a plurality of associated content source paths according to the influence coefficient of each behavior feature cluster, wherein the plurality of content source paths associated with each activity expression field are the content source paths of different field parameters of the activity expression field;

obtaining a content source path sequence of a plurality of content source paths associated with each activity expression field according to the influence coefficient of each activity expression field on the associated plurality of content source paths;

acquiring a plurality of candidate behavior feature clusters of candidate users;

acquiring content source path sequences of different content source paths corresponding to the candidate activity expression fields from the content source path sequences of the content source paths associated with the activity expression fields;

according to content source path sequences of different content source paths corresponding to the candidate activity expression fields, obtaining a content source path with the maximum influence of each field parameter of the candidate activity expression fields, obtaining target field parameters serving as the candidate activity expression fields based on the field parameters of the content source path with the maximum influence, and pushing content to the candidate user based on the target field parameters of each candidate activity expression field;

the step of obtaining the reference behavior expression network specifically includes:

acquiring a plurality of template behavior feature clusters from template user behavior data;

configuring a base behavior expression network by:

generating network members corresponding to the plurality of template behavior feature clusters;

determining a first class member connection relation in the basic behavior expression network between every two template behavior feature clusters which are matched with a first matching state, wherein the first matching state is the content source path matching of field parameters of template activity expression fields covered by every two template behavior feature clusters;

determining a second type of member connection relation in the basic behavior expression network between every two template behavior feature clusters which are matched in a second matching state, wherein the second matching state is that the template user behavior activity and the template activity expression field covered by every two template behavior feature clusters are matched;

determining influence coefficients corresponding to the plurality of template behavior feature clusters, and determining influence coefficients corresponding to the first class member connection relations and the second class member connection relations respectively;

loading the influence coefficients corresponding to the plurality of template behavior feature clusters, the first class member connection relations and the second class member connection relations in the basic behavior expression network, and determining a first loading network;

performing traversal optimization on the first loading network, and generating the reference behavior expression network according to generated first traversal optimization information, wherein the first traversal optimization information includes: and the influence coefficient of each template behavior characteristic cluster after traversal optimization.

2. The method for analyzing user behavior activity by using artificial intelligence analysis according to claim 1, wherein the step of determining the influence coefficients corresponding to the plurality of template behavior feature clusters specifically comprises:

taking an influence coefficient of a hot spot behavior feature cluster in the plurality of template behavior feature clusters as a first influence coefficient, wherein the influence coefficient of the hot spot behavior feature cluster is an influence coefficient of hot spot parameter statistics, the first influence coefficient is included in a first influence coefficient interval, the first influence coefficient interval is a threshold closed interval formed by a first threshold and a second threshold, and the first threshold is smaller than the second threshold;

and taking the influence coefficient of a non-hotspot behavior characteristic cluster in the plurality of template behavior characteristic clusters as a second influence coefficient, wherein the influence coefficient of the non-hotspot behavior characteristic cluster is the influence coefficient of hotspot parameter statistics, and the coefficient value of the second influence coefficient is the interval middle coefficient value of the first influence coefficient interval.

3. The method for analyzing user behavior activity by using artificial intelligence analysis according to claim 2, wherein the step of using an influence coefficient of a hotspot behavior feature cluster among the plurality of template behavior feature clusters as a first influence coefficient specifically comprises:

determining an influence coefficient of a template behavior feature cluster of a first target magnitude in the basic behavior expression network;

determining influence coefficients corresponding to the first class member connection relations and the second class member connection relations in the basic behavior expression network; wherein the first class member connection relationship is a member connection relationship between every two template behavior feature clusters matching the first matching state; the second class of member connection relation is the member connection relation between every two template behavior feature clusters which are matched with the second matching state;

loading the influence coefficients corresponding to the template behavior feature cluster of the first target magnitude, the first class member connection relation and the second class member connection relation in the basic behavior expression network, and determining a second loading network;

according to the influence coefficients corresponding to the template behavior feature clusters of the first target magnitude, the first class member connection relations and the second class member connection relations in the second loading network, performing traversal optimization on the influence coefficients of the template behavior feature clusters in the second loading network, and outputting second traversal optimization information, where the second traversal optimization information includes: influence coefficients of each template behavior feature cluster after traversal optimization;

taking the template behavior feature cluster with the influence coefficient larger than the set influence coefficient after traversal optimization as a hot spot behavior feature cluster;

and taking the influence coefficient of the hot spot behavior feature cluster as a first influence coefficient.

4. The method for analyzing user's behavioral activities according to claim 2, wherein before the step of using the influence coefficients of the hot behavior feature clusters of the plurality of template behavior feature clusters as the first influence coefficients, the method further comprises:

obtaining the template user behavior activity with the triggering frequency in the basic behavior expression network being greater than the set frequency, and obtaining the template user behavior activity as the hot spot user behavior activity;

acquiring one template behavior characteristic cluster from the plurality of template behavior characteristic clusters comprising each hotspot user behavior activity, and taking the acquired template behavior characteristic cluster as a hotspot behavior characteristic cluster corresponding to the hotspot user behavior activity;

and acquiring a plurality of hotspot influence coefficients corresponding to each hotspot behavior feature cluster, and taking the mean influence coefficient of the hotspot influence coefficients as the influence coefficient corresponding to the hotspot behavior feature cluster.

5. The method for analyzing user behavior and activity according to claim 2, wherein the step of determining the influence coefficients corresponding to the first class member connection relationships and the second class member connection relationships specifically includes:

obtaining a third influence coefficient by using the influence coefficient of the connection relationship between every two template behavior feature clusters matched with the first matching state, wherein the third influence coefficient is contained in a second influence coefficient interval, the second influence coefficient interval is a threshold value open interval formed by the second influence coefficient and a second threshold value, and the second influence coefficient is smaller than the second threshold value;

obtaining an influence coefficient of the connection relation of the second class members between every two template behavior feature clusters which are matched with the second matching state according to a heuristic global optimization algorithm;

the step of obtaining an influence coefficient of the connection relationship between every two template behavior feature clusters matching the second matching state according to a heuristic global optimization algorithm specifically includes:

determining an influence coefficient of the connection relation of the second class members between every two template behavior feature clusters based on a first strategy under the condition that the template activity expression fields covered by every two template behavior feature clusters are category-type template activity expression fields;

determining an influence coefficient of the connection relation of the second type of members between every two template behavior feature clusters based on a second strategy under the condition that the template activity expression fields covered by every two template behavior feature clusters are numerical template activity expression fields;

the step of determining an influence coefficient of the second-class member connection relationship between every two template behavior feature clusters based on the first policy specifically includes:

determining a first relevance cost value between field parameters of two template activity expression fields covered by every two template behavior feature clusters and a maximum hot spot coverage rate between the field parameters of the two template activity expression fields;

determining a product of the first relevance cost value and a third threshold value, determining a ratio of the product to the maximum coverage rate of the hot spot, wherein the third threshold value is larger than the second threshold value, and subtracting a difference value generated by the ratio from 1 to obtain a second relevance cost value between each two template behavior feature clusters;

obtaining the second relevance cost value between every two template behavior characteristic clusters to obtain an influence coefficient of the connection relation of the second class of members between every two template behavior characteristic clusters;

the step of determining an influence coefficient of the connection relationship between every two template behavior feature clusters of the second class member based on the second policy specifically includes:

obtaining an influence coefficient of the connection relation of the second type of members between every two template behavior feature clusters as a fourth influence coefficient under the condition that field parameters of two template activity expression fields covered by every two template behavior feature clusters are matched, wherein the coefficient value of the fourth influence coefficient is the second threshold value;

and under the condition that the field parameters of the two template activity expression fields covered by every two template behavior feature clusters are not matched, obtaining that the influence coefficient of the connection relationship of the second class members between every two template behavior feature clusters is a fifth influence coefficient, and the coefficient value of the fifth influence coefficient is the first threshold value.

6. The method for analyzing user behavior and activity by using artificial intelligence analysis according to claim 1, wherein the step of performing traversal optimization on the first loading network specifically comprises:

obtaining the influence coefficient of each template behavior feature cluster after traversal optimization based on the following steps during each traversal optimization:

determining a plurality of connected template behavior characteristic clusters corresponding to the template behavior characteristic clusters in the first loading network, and determining the product of the influence coefficients corresponding to the connected template behavior characteristic clusters; the influence coefficient product corresponding to each connected template behavior feature cluster is obtained by multiplying the following elements: influence coefficients of the connected template behavior feature clusters and influence coefficients of member connection relations between the connected template behavior feature clusters and the template behavior feature clusters; under the condition that the connected template behavior feature cluster and the template behavior feature cluster are matched in the first matching state, the influence coefficient of the member connection relationship is the influence coefficient corresponding to the first class of member connection relationship, and under the condition that the connected template behavior feature cluster and the template behavior feature cluster are matched in the second matching state, the influence coefficient of the member connection relationship is the influence coefficient corresponding to the second class of member connection relationship;

and adding the products of the influence coefficients corresponding to the plurality of connected template behavior feature clusters to obtain the influence coefficients optimized for the traversal of the template behavior feature clusters.

7. The method for analyzing user behavioral activities according to artificial intelligence analysis according to claim 1, wherein the step of determining the reference behavior expression network according to the generated first traversal optimization information specifically includes:

in the plurality of template behavior characteristic clusters comprising the same template user behavior activity and template activity expression fields, traversing the optimized template behavior characteristic cluster with the largest influence coefficient to obtain a first target behavior characteristic cluster corresponding to the template user behavior activity;

and determining the reference behavior expression network according to the first target behavior feature cluster corresponding to each template user behavior activity covered by the first loading network and the influence coefficient corresponding to each first target behavior feature cluster.

8. The method for analyzing user behavior and activity by using artificial intelligence analysis according to claim 1, wherein the step of obtaining an influence coefficient of each activity expression field on the associated content source paths according to the influence coefficient of each behavior feature cluster specifically comprises:

and walking the activity expression field, and walking each content source path associated with the activity expression field to which the current walking is carried out, and carrying out the following operations:

determining a first behavior feature cluster in the reference behavior expression network, the first behavior feature cluster including the activity expression field to which the user currently walks;

determining a second behavior feature cluster and a second target magnitude of the second behavior feature cluster, wherein the second behavior feature cluster comprises the activity expression field which is currently walked to, and the corresponding field parameter comes from the content source path which is currently walked to;

obtaining a ratio of the sum of the influence coefficients of the second behavior feature clusters to the second target magnitude to obtain an influence coefficient of the activity expression field currently walked to on the content source path currently walked to;

the step of obtaining a content source path sequence of the plurality of content source paths associated with each activity expression field according to the influence coefficient of each activity expression field on the associated plurality of content source paths specifically includes:

for each of the activity expression fields, performing the following:

aiming at a plurality of content source paths related to the activity expression field, obtaining an influence coefficient of the activity expression field on each content source path;

sorting the influence coefficients of the activity expression fields on the plurality of content source paths in a descending order, and outputting a content source path sequence of the activity expression fields on the plurality of content source paths;

after obtaining the target field parameter of the candidate activity expression field based on the field parameter of the content source path with the maximum influence, the method further includes:

taking the candidate behavior feature cluster comprising the target field parameter as a second target behavior feature cluster;

loading the second target behavior feature cluster serving as a new network member to the reference behavior expression network, and connecting the second target behavior feature cluster with a third behavior feature cluster covered in the reference behavior expression network to generate a third behavior expression network; the user behavior activities covered by the third behavior feature cluster are matched with the user behavior activities covered by the second target behavior feature cluster, or the content source paths of the field parameters of the covered activity expression fields are matched with the content source paths of the field parameters of the activity expression fields covered by the second target behavior feature cluster;

after the generating the third behavioral expression network, the method further comprises:

obtaining respective corresponding influence coefficients of a plurality of behavior feature clusters, a first class member connection relation and a second class member connection relation in the third behavior expression network under the condition that data synchronization is not performed, wherein the condition that data synchronization is not performed is a specific state of a content pushing process of a response aiming at the candidate behavior feature cluster; the first class of member connection relation is the member connection relation between every two behavior feature clusters which are matched with a first matching state, and the first matching state is the content source path matching of field parameters of activity expression fields covered by every two behavior feature clusters; the second type of member connection relation is a member connection relation between every two behavior feature clusters which matches a third requirement, and the third requirement is the matching of user behavior activities covered by every two behavior feature clusters;

loading the influence coefficients corresponding to the plurality of behavior feature clusters, the first class member connection relations and the second class member connection relations in the third behavior expression network, and determining a third loading network;

performing traversal optimization on the third loading network, and determining the optimized reference behavior expression network according to generated third traversal optimization information, where the third traversal optimization information includes: and the influence coefficient of each behavior feature cluster after traversal optimization.

9. A big data system, comprising a processor, a machine-readable storage medium, and a network interface, wherein the machine-readable storage medium, the network interface, and the processor are associated with each other through a bus system, the network interface is configured to be communicatively connected to a plurality of cloud service servers, the machine-readable storage medium is configured to store a program, an instruction, or code, and the processor is configured to execute the program, the instruction, or code in the machine-readable storage medium to perform the method for analyzing user behavior activity by using artificial intelligence analysis according to any one of claims 1 to 8.