E-commerce Personalized Recommender System Research

Web2.0era, the e-commerce has developed rapidly. But with the development ofthe electronic commerce the "information overload" problem has brought out inevitably.The personalized recommendation system is used to solve this problem. ThePersonalized Recommendation Systems for E-Commerce is the use of e-commerce siteto provide customers with information and advice to help users decide to purchaseproduct, and to simulate sales staff help customers to complete the purchase process.The high-quality personalized recommendation system has brought huge economic andsocial values, but also can improve the viability of e-commerce site in today’s highlycompetitive market environment.This paper describes several recommendation algorithms used in the personalizedrecommendation system, such as: collaborative filtering recommendation algorithms,content-based recommendation algorithm, the grid-based recommendation algorithmand so on, and then describe and analyze their advantages and disadvantages. Thecollaborative filtering algorithm is the oldest and most widely used recommendationalgorithm today. The problems in the traditional collaborative filtering recommendationalgorithms: similarity calculation does not consider the difference between a single userpoint of interest, user jointly scoring too little and new items, the improved algorithm isproposed. Improved collaborative filtering algorithm uses the relativity of the items toamend the Pearson coefficient, and then the amended Pearson coefficient is combinedwith the conditional probability of the user common score linearly to get the finalsimilarity. For new projects, through the project in project-based collaborative filteringrecommendation algorithm based on correlation calculations to fix the originalproject-based collaborative filtering algorithm Unable to get the most adjacent item setsthe issue for the new project. Then verify the improved collaborative filteringalgorithms and a new project solutions in the MovieLens data set, compared with thetraditional collaborative filtering recommendation algorithm accuracy obtainedimproved collaborative filtering algorithm does improve the accuracy of therecommended results, in addition, by solution calculate the new project in the MAE of MovieLens data set, it is worth, the solution to solve the problem of new projects to acertain extent, the solution obtained in this paper is effective and feasible.The improved collaborative filtering algorithm used in a buy-platform project,through describing the design and implementation of recommended modules, andultimately achieve a practical purpose.