CN-122022950-A - Value-added service generation method based on behavior characteristics

CN122022950ACN 122022950 ACN122022950 ACN 122022950ACN-122022950-A

Abstract

The invention discloses a value added service generation method based on behavior characteristics, which particularly relates to the technical field of service recommendation, and is characterized in that a behavior characteristic initial data set is generated by acquiring owner consumption behavior data in the source service field and the target service field, characteristic distribution differences of owner consumption behaviors in the cross-field are identified by utilizing a cluster analysis method, the cross-field characteristic association direction is determined based on correlation analysis, reverse migration characteristics in negative association are identified, a characteristic optimization data set is obtained by giving differential weight to negative influences of the reverse migration characteristics on target field recommendation accuracy, then a field self-adaptive recommendation model is constructed, negative migration influences of the source service field characteristics in the target field are reduced by utilizing countermeasure training, and cross-field behavior characteristic expression is obtained, so that an owner personalized value added service recommendation scheme suitable for the target service field is generated, and recommendation accuracy and user experience are improved.

Inventors

HUANG ZEYUAN
CHEN TONG
SONG XIAOHUI

Assignees

青民数科(青岛)技术服务有限公司

Dates

Publication Date: 20260512
Application Date: 20260129

Claims (7)

1. The value-added service generation method based on the behavior characteristics is characterized by comprising the following steps of: s1, acquiring owner consumption behavior data in the source business field and the target business field, and preprocessing to generate an initial behavior characteristic data set; S2, performing cluster analysis on the behavior feature initial data set, and identifying feature distribution differences of the consumption behaviors of the vehicle owners in the source service field and the target service field to generate a difference feature set; S3, based on the difference feature set, determining the association direction between the consumption behavior features of the vehicle owners in the source service field and the target service field through correlation analysis, and identifying reverse migration features with negative feature association; S4, according to the negative influence degree of the reverse migration feature on the accuracy of the recommendation result in the target service field, giving a difference weight to the corresponding feature in the initial data set of the behavior feature to obtain a feature optimization data set; s5, constructing a domain self-adaptive recommendation model according to the feature optimization data set, and reducing negative migration influence of source service domain features in the target service domain by using an countermeasure training method to obtain cross-domain behavior feature expression; s6, generating a personalized value-added service recommendation scheme of the vehicle owner applicable to the target service field based on the cross-field behavior feature expression.
2. The value-added service generation method based on behavior characteristics according to claim 1, wherein the method is characterized in that the method comprises the steps of obtaining the consumption behavior data of the vehicle owners in the source service field and the target service field, preprocessing the consumption behavior data, and generating an initial data set of the behavior characteristics, wherein the initial data set of the behavior characteristics is specifically as follows: Collecting original data of the consumption behavior of a vehicle owner in the source service field and the target service field; removing invalid records and supplementing missing values from the original data; and carrying out standardized conversion on the original data according to the unified data format specification to obtain an initial data set of the behavior characteristics.
3. The value added service generation method based on behavior characteristics according to claim 2, wherein the clustering analysis is performed on the initial data set of behavior characteristics, and the feature distribution differences of the consumption behaviors of the owners in the source service field and the target service field are identified, so as to generate a difference feature set, which specifically comprises: Forming a domain subset from the behavior feature initial data set according to the source service domain and the target service domain respectively; selecting the consumption amount characteristics, the consumption frequency characteristics and the class preference characteristics of the field subset, and performing cluster analysis to obtain a field cluster label and a field cluster center; And performing similarity matching on the source service field cluster center and the target service field cluster center, and calculating the characteristic distribution divergence in the field clusters to obtain a difference characteristic set.
4. The value added service generation method based on behavior characteristics according to claim 3, wherein the correlation direction between the consumption behavior characteristics of the vehicle owner in the source service field and the target service field is determined through correlation analysis based on the difference characteristic set, and the identification characteristic correlation is a negative reverse migration characteristic, specifically: Extracting feature sequences corresponding to the difference feature sets from the source service field subset and the target service field subset based on the difference feature sets; Grouping the feature sequences according to the domain cluster labels and calculating feature correlation coefficients between a source service domain cluster center and a target service domain cluster center; And converging the characteristics with the characteristic correlation coefficient smaller than zero and stable matching between the source service field cluster label and the target service field cluster label to obtain the reverse migration characteristics.
5. The method for generating value-added service based on behavioral characteristics according to claim 4, wherein the method for generating value-added service based on behavioral characteristics is characterized in that according to the negative influence degree of reverse migration characteristics on the accuracy of the recommended result in the target service field, a difference weight is given to the corresponding characteristics in the initial dataset of behavioral characteristics, and a feature optimization dataset is obtained, specifically: extracting feature values corresponding to the reverse migration features from the behavior feature initial data set based on the reverse migration features and forming a reverse migration feature subset; calculating the center deviation degree of the domain cluster of the reverse migration feature in the target service domain subset for the reverse migration feature subset and forming a deviation degree set; Mapping the reverse migration feature subset into suppression weights based on the deviation degree set, and backfilling the suppression weights to the behavior feature initial data set to obtain a feature optimization data set.
6. The method for generating value-added service based on behavior feature according to claim 5, wherein a domain self-adaptive recommendation model is constructed according to a feature optimization data set, negative migration influence of source service domain features in a target service domain is reduced by using an countermeasure training method, and cross-domain behavior feature expression is obtained, specifically: Respectively forming a training sample set according to the source service field and the target service field by the feature optimization data set, and constructing training pairs of behavior features and value-added service items; Training the field discrimination parameters of the field self-adaptive recommendation model based on the recommended prediction parameters of the training field self-adaptive recommendation model based on the training and the field labels of the training sample set; And adopting countermeasures to introduce output difference inverse constraint of domain discrimination parameters and keep domain discrimination training of the domain discrimination parameters when updating recommended prediction parameters, and obtaining cross-domain behavior feature expression.
7. The value-added service generation method based on the behavior characteristics according to claim 6, wherein the generation of the personalized value-added service recommendation scheme of the vehicle owner applicable to the target service field is based on the cross-domain behavior characteristic expression, specifically: calculating the matching score of each value added service item in the target service field based on the cross-field behavior feature expression and forming a candidate value added service item set; Sorting the candidate value added service item sets according to the matching scores and executing duplicate removal and coverage constraint screening by combining with the target service field cluster labels to obtain value added service item combinations; and converting the value added service item combination into an executable value added service recommendation scheme in the target service field.

Description

Value-added service generation method based on behavior characteristics Technical Field The invention relates to the technical field of service recommendation, in particular to a value added service generation method based on behavior characteristics. Background In the car owner service operation scene, service bodies such as car dealers, insurance companies, car finance companies and the like usually develop value-added service recommendation and combined configuration based on car owner consumption behavior data to form a service recommendation scheme facing to the target service field, and because of the difference between service data sources and service supply forms, the prior art often needs to use car owner behavior characteristics of the source service field for service matching of the target service field so as to realize value-added service generation and landing across the service field. In the prior art, when a value-added service recommendation scheme is generated by utilizing the consumption behavior characteristics of an owner in a cross-service domain, the suitability of the behavior characteristics of a source service domain in a target service domain is insufficient, so that a service matching result obtained based on the behavior characteristics of the source service domain is deviated in the target service domain. Disclosure of Invention In order to overcome the above-mentioned drawbacks of the prior art, an embodiment of the present invention provides a value added service generation method based on behavioral characteristics to solve the above-mentioned problems set forth in the background art. In order to achieve the above purpose, the present invention provides the following technical solutions: A value added service generation method based on behavior characteristics comprises the following steps: s1, acquiring owner consumption behavior data in the source business field and the target business field, and preprocessing to generate an initial behavior characteristic data set; S2, performing cluster analysis on the behavior feature initial data set, and identifying feature distribution differences of the consumption behaviors of the vehicle owners in the source service field and the target service field to generate a difference feature set; S3, based on the difference feature set, determining the association direction between the consumption behavior features of the vehicle owners in the source service field and the target service field through correlation analysis, and identifying reverse migration features with negative feature association; S4, according to the negative influence degree of the reverse migration feature on the accuracy of the recommendation result in the target service field, giving a difference weight to the corresponding feature in the initial data set of the behavior feature to obtain a feature optimization data set; s5, constructing a domain self-adaptive recommendation model according to the feature optimization data set, and reducing negative migration influence of source service domain features in the target service domain by using an countermeasure training method to obtain cross-domain behavior feature expression; s6, generating a personalized value-added service recommendation scheme of the vehicle owner applicable to the target service field based on the cross-field behavior feature expression. In a preferred embodiment, the method comprises the steps of obtaining the consumption behavior data of the vehicle owners in the source business field and the target business field, preprocessing the consumption behavior data, and generating an initial behavior characteristic data set, wherein the initial behavior characteristic data set specifically comprises the following steps: Collecting original data of the consumption behavior of a vehicle owner in the source service field and the target service field; removing invalid records and supplementing missing values from the original data; and carrying out standardized conversion on the original data according to the unified data format specification to obtain an initial data set of the behavior characteristics. In a preferred embodiment, the initial data set of behavior features is subjected to cluster analysis, and feature distribution differences of the consumption behaviors of the owners in the source service field and the target service field are identified to generate a difference feature set, which specifically includes: Forming a domain subset from the behavior feature initial data set according to the source service domain and the target service domain respectively; selecting the consumption amount characteristics, the consumption frequency characteristics and the class preference characteristics of the field subset, and performing cluster analysis to obtain a field cluster label and a field cluster center; And performing similarity matching on the source service field cluster center and the target service