Research On Payment Behavior Recognition For Mobile Users Based On A Knowledge-and-Data-Driven Approach

With the rapid development of mobile payment,the scale of payment data is exploding.Businesses are eager to use users' consumption behavior preferences to quickly locate the target consumer groups.The effective mining of payment behavior data is necessary to analyze users' consumption preferences.This thesis,based on the “Big Data Public Data Model Development Service” project supported by the enterprise,uses mobile data to analyze the users' payment behavior preferences,including the payment modes,time and location.This thesis proposes a novel knowledge-and-data-driven modeling approach for mobile users' payment behavior recognition to improve the accuracy and efficiency of traditional method.Before identifying the payment behavior,the payment events identification is introduced.The data-driven model is used to identify behavioral events and filter payment events,then the knowledge-driven model is used to identify the payment behavior.In the aspect of data-driven model,aiming at the inaccuracy of online behavior segmentation,this thesis proposes an event-driven approach of segmentation to divide online behavior by using the event-driven characteristics,which improve the effectiveness of model feature extraction.Then this thesis extracts payment-related features from segmented behavior events,and uses the data-driven model to identify the payment events.In the aspect of knowledge-driven model,aiming at the defects of the keyword extraction method and the lack of knowledge of the mobile payment field,this thesis proposes a keyword extraction approach of mobile payment modes based on improved TF-IDF,and constructs the mobile payment modes keyword repository in knowledge-driven to improve the recognition performance of the mobile payment modes.At the same time,based on the location information of the base station in the mobile signaling data,this thesis constructs the user behavior location repository to realize mobile payment time and location identification.The experimental results show that the event-driven approach of segmentation method is higher in accuracy and recall than the comparison method for identifying the mobile payment events.In the aspect of knowledge-and-data-driven model,the F1 value of the proposed approach is more superior than others,and recognition accuracy of Alipay and WeChat payment is 88.3% and 84.5% respectively,and no payment is 95.5%.