CN-122022862-A - Operator user portrait construction method and system based on rules and vector features

CN122022862ACN 122022862 ACN122022862 ACN 122022862ACN-122022862-A

Abstract

The invention discloses an operator user portrait construction method and system based on rules and vector features, wherein the method comprises the steps of obtaining desensitized user data based on an operator database and processing the desensitized user data to obtain a structured feature vector, matching the structured feature vector based on a preset rule base to obtain a rule hit score, constructing a graphic neural network based on the structured feature vector, mapping the vector features into vector feature scores after determining vector features of user nodes in the graphic neural network, inputting the rule hit score and the vector feature scores into a gating fusion network to obtain fusion scores, carrying out feature analysis and rule path tracking based on the fusion scores to determine rule hit information and feature contribution, writing the fusion scores, rule hit information, abstract of the vector features and the feature contribution into a user portrait, and improving accuracy of the operator user portrait construction process and generalization capability of the user portrait construction process.

Inventors

ZHAO YAHUI
Li Zhangti
ZHANG YAWEI
MA YU
QI ZHENFENG
ZHOU BISHU
WU RUIQI
TANG ZHONGYANG
ZHOU LEI

Assignees

中国联合网络通信有限公司软件研究院

Dates

Publication Date: 20260512
Application Date: 20251224

Claims (8)

1. The operator user portrait construction method based on the rule and the vector features is characterized by comprising the following steps: acquiring desensitized user data based on an operator database, and processing the user data to obtain a structured feature vector; Matching the structured feature vectors based on a preset rule base to obtain a rule hit score, constructing a graph neural network based on the structured feature vectors, and mapping the vector features into vector feature scores after determining the vector features of user nodes in the graph neural network; inputting the rule hit score and the vector feature score into a gating fusion network to obtain a fusion score; performing feature attribution analysis and rule path tracking based on the fusion score to determine rule hit information and feature contribution; and writing the fusion score, the rule hit information, the abstract of the vector feature and the feature contribution into the user portrait.
2. The method for constructing the operator user portrayal based on the rule and the vector feature according to claim 1, wherein the rule hit score is obtained by matching the structured feature vector based on a preset rule base, specifically, the rule hit score is obtained by the following formula: ; In the formula, For a rule hit score, As the weight of the kth rule, In order to be a number of rules, As a result of the hit of the kth rule, Is a bias term.
3. The method for constructing operator user portrayal based on rule and vector features according to claim 1, wherein the constructing of the graph neural network based on the structured feature vector is to construct a graph structure by taking an entity related to a user as a node and taking a behavior related to the user as an edge, wherein a time stamp is recorded in the edge.
4. The operator user portrayal construction method based on rules and vector features of claim 3, characterized in that the vector features are mapped into vector feature scores by the following formula: ; In the formula, As a score of the vector feature, As a weight vector of the weight vector, For the purpose of the transposition, As a feature of the vector it is, Is a bias term.
5. The operator user portrayal construction method based on rules and vector features of claim 1, wherein the fusion score is determined specifically by the following formula: ; ; In the formula, In order to fuse the scores, the score, In order to fuse the weights, the weights are, For a rule hit score, As a score of the vector feature, In order to gate the weight matrix, In order to splice the rule vectors and the vector features to obtain a spliced vector, As a result of the bias term, To vectorize rules into a rule vector, Is a vector feature.
6. The method for constructing a customer representation of an operator based on rules and vector features according to claim 5, wherein the feature-attribution analysis is performed on the fused score, the features including rules, vector features and key behavior features, and the feature-attribution analysis is performed to determine feature contribution by the following formula: ; ; In the formula, The feature contribution for the ith feature, In order to not include a subset of the ith feature, In order to contain the set of all the features, In order to output a function of the fusion score, Is a baseline term.
7. The method for constructing an operator user portrayal based on rule and vector features of claim 6, wherein said performing rule path tracking on the fused score comprises: a rule based on fusion score query hits; determining the feature contribution degree of the hit rule and the original feature, rule condition, rule label and fusion score corresponding to the hit rule, and combining the feature contribution degree, the original feature, the rule condition, the rule label and the fusion score corresponding to the hit rule into a tracking path; and saving the tracking path into rule hit information corresponding to the hit rule.
8. An operator portrayal construction system based on rules and vector features, said system comprising: The acquisition module is used for acquiring the desensitized user data based on the operator database and processing the user data to obtain a structural feature vector; The score module is used for matching the structured feature vectors based on a preset rule base to obtain a rule hit score, constructing a graph neural network based on the structured feature vectors, and mapping the vector features into vector feature scores after determining the vector features of the user nodes in the graph neural network; the fusion module is used for inputting the rule hit score and the vector feature score into a gating fusion network to obtain a fusion score; The determining module is used for carrying out feature attribution analysis and rule path tracking based on the fusion score to determine rule hit information and feature contribution degree; and the portrait module is used for writing the fusion score, the rule hit information, the abstract of the vector feature and the feature contribution into the user portrait.

Description

Operator user portrait construction method and system based on rules and vector features Technical Field The invention belongs to the technical field of communication, and particularly relates to an operator user portrait construction method and system based on rules and vector features. Background The user portraits, namely user information tagging, are significant to telecom operators in establishing accurate user portraits for carrying out stock management user maintenance and value improvement, and can effectively improve the efficiency and quality of product marketing and service. The existing operator user portrait construction scheme depends on a rule tag system, has clear rule logic and insufficient generalization capability, and when a new mode beyond a preset rule appears, the new mode specifically refers to the situation that the existing rule base is difficult to cover in time to cause the degradation of user portrait accuracy due to the fact that a service strategy is adjusted, a product form is changed, a user behavior combination rule is changed due to application entrance change or user behavior migration, a cross-entity relation structure is changed or relation strength distribution is changed, and the existing technical scheme is difficult to adapt to the new mode. Therefore, how to improve accuracy of operator user portraits and generalization capability of user portraits in construction process is a technical problem to be solved by those skilled in the art. Disclosure of Invention The invention aims to solve the technical problems of low accuracy of operator user portrayal and low generalization capability in the construction process in the prior art. In order to achieve the technical purpose, in one aspect, the invention provides a method for constructing an operator user portrait based on rules and vector features, the method comprising: acquiring desensitized user data based on an operator database, and processing the user data to obtain a structured feature vector; Matching the structured feature vectors based on a preset rule base to obtain a rule hit score, constructing a graph neural network based on the structured feature vectors, and mapping the vector features into vector feature scores after determining the vector features of user nodes in the graph neural network; inputting the rule hit score and the vector feature score into a gating fusion network to obtain a fusion score; performing feature attribution analysis and rule path tracking based on the fusion score to determine rule hit information and feature contribution; and writing the fusion score, the rule hit information, the abstract of the vector feature and the feature contribution into the user portrait. Further, the structural feature vector is matched based on a preset rule base to obtain a rule hit score, and the rule hit score is obtained specifically through the following formula: ; In the formula, For a rule hit score,As the weight of the kth rule,In order to be a number of rules,As a result of the hit of the kth rule,Is a bias term. Further, the graph neural network is constructed based on the structural features, specifically, entities related to users are taken as nodes, behaviors related to the users are taken as edges, and a graph structure is constructed, wherein time stamps are recorded in the edges. Further, the vector features are mapped into vector feature scores specifically by the following formula: ; In the formula, As a score of the vector feature,As a weight vector of the weight vector,For the purpose of the transposition,As a feature of the vector it is,Is a bias term. Further, the fusion score is specifically determined by the following formula: ; ; In the formula, In order to fuse the scores, the score,In order to fuse the weights, the weights are,For a rule hit score,As a score of the vector feature,In order to gate the weight matrix,In order to splice the rule vectors and the vector features to obtain a spliced vector,As a result of the bias term,To vectorize rules into a rule vector,Is a vector feature. Further, the feature attribution analysis is performed on the fusion score, wherein the features comprise rules, vector features and key behavior features, and the feature attribution analysis is performed through the following formula to determine feature contribution degree: ; ; In the formula, The feature contribution for the ith feature,In order to not include a subset of the ith feature,In order to contain the set of all the features,In order to output a function of the fusion score,Is a baseline term. Further, the step of performing regular path tracking on the fusion score specifically includes: a rule based on fusion score query hits; determining the feature contribution degree of the hit rule and the original feature, rule condition, rule label and fusion score corresponding to the hit rule, and combining the feature contribution degree, the original feature, the rule condition, the