Application And Research Of User Behavioral Analysis Technology In Personalized Service

This paper mainly studies the application of user behavior analysis and personalized service technology in the field of e-commerce.The e-commerce platform can use user behavior analysis technology to analyze user behavior characteristics and provide personalized information services,that is personalized recommendation,according to the analysis results.This paper chooses neural network technology and association rule technology to analyze user behavior,and designs the process of the personalized service based on the analysis results by using collaborative filtering algorithm.An improved Pearson similarity method is designed to solve the problem that When using user-based collaborative filtering algorithm to calculate user similarity with Pearson similarity algorithm,the similarities result in bias.Considering the two factors of active users and hot items,the contribution of active users and hot items to the similarity calculation is reduced,so that the calculation of similarity between users is more accurate,and the algorithm can improve the mining ability of long tail items and obtain better recommendation effects.To the data of user data,product data and behavior date in e-commerce,and calculate the relationship between users and items by neural network to construct a new neural network model for e-commerce data.Screening candidate set by Hybrid improved collaborative filtering algorithm and filling it by association rules between categories of products.Using neural network to predict the score of candidate set,in order to improve accuracy and diversity of recommend results.In order to improve the cold start problem of user-based collaborative filtering algorithm,the similarity calculation method based on users'static attributes is used to calculate the similarity between new users and old users,and the recently purchased items of old users are recommended to new users.Finally,the paper analyzes the data characteristics of the e-commerce platform,designs and implements the personalized recommendation system,and conducts experiments and results analysis on the personalized recommendation function.The experimental results show that the improved collaborative filtering algorithm can improve the computing accuracy and recommendation quality of similar users.The hybrid recommendation algorithm which solves the cold start problem of users to a certain extent,improves the accuracy and diversity of recommendation,and improves the quality of recommendation.