Study On The Recommendation System Of Dynamic User Preference Model Based

The rapid development of the Internet,Web services,and Web-based information systems has led to a surge in network resources.Recommendation has become an indispensable and powerful tool to screen information for Internet users under the background of information overload.At the same time,the scale of users is also rapidly increasing.The proportion of network users using e-commerce is increasing,and the recommendation system is gradually improving.In the field of e-commerce,it has made extraordinary achievements by providing users with accurate marketing and perfecting website services.The goal of e-commerce recommendation system is to improve the click-through rate and order conversion rate.Although many scholars have carried out in-depth research on the direction of recommendation algorithm,there are still some problems.The algorithm lacks the ability to capture the user's needs and it's dynamic and migratory nature.In addition,the user's purchase behavior has a Periodicity.The system lacks analysis of this nature.The e-commerce platforms recommend the user with newly purchased or recently purchased duplicate commodities,causing poor user experience and user disgust.In order to solve the above problems,this paper studies and implements a recommendation system for tracking user interest drift based on the Mingxi Journey to E-commerce platform.The research content includes the following points:first,On the basis of understanding of the data features and manifestations of e-commerce transaction logs and user information,the recommendation algorithm with user behavior time could emphasize the change of user group interest with the advancement of time.Secondly,according to the phenomenon of user interest drift,the user interest models are divided into short-term and long-term categories,and Forgetting curve is introduced to construct the user interest weight attenuation function to simulate the movement of user interest preferences.Thirdly,according to the application scenario of e-commerce platform,commodity categories are abstractly used to represent users' purchase interest points,and a user-category matrix is constructed to improve the phenomenon of similarity deviation caused by sparse matrix in collaborative recommendation.Fourthly,analyzes the periodic characteristics of e-commerce users'purchase behavior,and combines the concept of user repurchase rate and commodity repurchase cycle to design a recommended correction model for introducing commodity repurchase cycle.The obtained correction factor is the recommended candidate set for the target user.Fifth,build and implement Minglai Journey to the West e-commerce platform based on user dynamic interest model recommendation system,give the overall architecture design of the platform,and implement the core functions of each module based on SSM architecture mode.At last,the test results show that the user interest model can accurately express the distribution of user interest points,and the recommendation algorithm based on the user interest model improves the recommendation recall rate.By comparing the recommendation effect of the dynamic model with that of the static model,the results show that the dynamic model has higher accuracy,which verifies the effectiveness of updating based on the Forgetting curve model.