Research On Personalized Recommendation Based On User Behavior

Since the twenty-first Century,with the rapid development of Internet and electronic commerce,mass commodity information has gradually become a constraint on the purchase of goods by network consumers.Therefore,personalized recommendation system built on massive data mining has been developed rapidly.Not only many electric business platforms have adopted personalized intelligent recommendation system,but also personalized recommendation system has gradually become a system.It is a hot issue in academic research.Accurate recommendation can effectively solve the problem of information overload and promote the sale of platform goods.Personalized recommendation algorithm based on collaborative filtering and personalized recommendation algorithm based on content are the two main research ideas of personalized recommendation.Meanwhile,the implicit feedback recommendation algorithm based on user historical behavior data is also widely studied.This paper first introduces some of the main classical algorithms in personalized recommendation research,and then analyzes the shortage of personalized recommendation research on user behavior at the present stage,and studies the personalized recommendation based on user historical behavior.In view of the demand of intelligent e-commerce platform,the algorithm model based on user behavior data analysis is constructed by analyzing user's historical behavior data,analyzing user behavior habit and mining user's commodity attribute preference.Through the research,we mainly realize the user classification based on the historical behavior of long term user,build the commodity brand preference system based on historical interactive information,build the commodity attribute preference system based on the user purchase and user browsing behavior.Based on the purchase intention forecast of the user's recent browsing,it realizes the short-term purchase demand forecast and the intentional commodity recommendation.Finally,the prediction model is verified by experimental data,indicating the feasibility of the recommendation algorithm and the direction to be improved.