Design And Implementation Of E-commerce Recommendation System Based On Hybrid Collaborative Filtering

With the continuous penetration of the Internet into people's daily life,the degree of informationization of people's daily life is increasing.While people obtain valuable information on the network,the flood of worthless information also seriously interferes with people's daily life.Especially among e-commerce websites,the amount of information on the website is generally around one million.The practical value of the recommendation system is revealed here.At present,China's more famous large-scale e-commerce sites,Jingdong and Taobao,both use recommendation systems to help users solve the problem of information overload.The recommendation system of e-commerce speculates the user's personal preference by collecting the user's daily behavior data,and uses this as a basis to implement the product recommendation function.However,the products recommended by many recommendation systems are not strictly consistent with the user's preferences.The low accuracy of the recommendation results of the recommendation system has also become the main problem faced by the recommendation system.In order to solve this problem and improve the accuracy of the recommendation results of the recommendation system,many studies have proposed a variety of improved algorithms based on the traditional recommendation system to improve the accuracy of the recommendation results.On this basis,this paper proposes an improved collaborative filtering recommendation system,which optimizes the calculation of the missing data of the scoring matrix,and uses the user 's historical browsing data as an important criterion for judging the user 's preferences.Among them,the proximity and frequency attributes as an indicator,calculating a high-quality scoring matrix can solve the problem of data sparsity,no longer need to set the noise weight,and the sensitivity to interference is relatively low.At the same time,this paper studies the dynamic scoring behavior,adds dynamic regularization factors and neighborhood factors to the collaborative filtering based on matrix decomposition,and updates the weights of the user feature vector and commodity feature vector according to the real-time changes of user behavior,improving the dynamic update of the recommendation system.Ability to simultaneously merge neighborhood factors to track the drift of user preferences to improve the accuracy of recommendations and increase user satisfaction.