Analysis And Research On E-commerce User Behavior Based On Implicit Feedback Recommendation Algorithm

With the rapid updating and iteration of Internet technology and the promotion and popularization of 5G network,the data on the Internet is growing explosively,while the cost of obtaining useful information is increasing.Various e-commerce platforms record a large number of implicit feedback data of users(such as clicking,collecting,adding to the shopping cart,etc.).Although they can not directly reflect the degree of user preferences,they also imply user preferences to a certain extent.Bayesian Personalized Ranking(BPR)is one of the most representative algorithms for implicit feedback problems,but the independence assumption between users and the pairwise preference assumption of individuals for two items are too strict.Based on the implicit feedback recommendation algorithm,this paper analyzes and studies the user behavior of e-commerce,finds the user’s interest from the implicit feedback data,and predicts and recommends products to users.The specific work is as follows:(1)An implicit feedback preference quantification model is introduced,and an enhance-Difference Personalized Ranking algorithm based on user behavior feedback is proposed.In the e-commerce platform,different types of behaviors of users have different implicit preferences,so we analyze this and construct an implicit feedback quantitative preference model to uniformly quantify and predict the probability value of deterministic feedback of users,which to some extent reflects the degree of preference of users for commodities in the future,and then introduce the prediction results into the Difference Personalized Ranking algorithm for improvement and optimization.Clicks are no longer used as the criterion of popularity.Finally,by comparing the new algorithm with the baseline algorithm on the same data set,it is verified that the new algorithm has better recommendation performance.(2)Improve the sampling strategy of users and commodities,and propose a Group Difference Personalized Ranking algorithm based on commodity content.Group preference is introduced into the preference difference algorithm.Users are grouped according to their consumption level and gender,and then randomly sampled to create more representative user groups,so as to relax the assumption of independence among users;The commodity sampling method in the preference difference algorithm is improved by using the commodity category information,and the triple sample formed by randomly selecting two commodities of the same category is more reliable than the triple sample formed by randomly selecting commodities;finally,the comparative experiment between the new algorithm and the baseline algorithm is carried out on the same data set,and the results show that the recommendation performance of the new algorithm is better.(3)Finally,the enhanced-Group Difference Personalized Ranking algorithm based on user behavior feedback and commodity content proposed in this paper is compared with other baseline algorithms in the experimental results of various evaluation indicators in Jingdong e-commerce data set.The results show that the algorithm proposed in this paper can improve the recommendation performance to a certain extent and recommend better results for users.