Research On E-commerce Recommendation Technologies Based On Web Mining

With the popularization of Internet and the development of E-commerce, the recommendation system is playing a more and more important role in E-commerce system. It can enhance E-commerce sales by converting browsers into buyers, increasing cross-sell and building loyalty to prevent user losing. But with the further expansion of the E-commerce, recommendation system is facing a series of challenges. Aimed at the main challenges of recommendation systems in E-commerce, the recommendation technologies in E-commerce based on Web mining are explored and researched on in this thesis.This thesis elaborates the technologies of Web mining and its applications in E-commerce recommendation system and the collaborative filtering recommendation algorithms and the recommendation algorithm based on association rule. Considering the problems such as the low precision, the algorithms in Web access sequential patterns mining and the recommendation results producing algorithm are improved in this thesis.Firstly, in the patterns mining stage, the improved FLWAP-mine algorithm is proposed based on WAP-tree. The algorithm constructs the tree by scanning the database just once and uses the ideas of projection tree and pruning to decrease the search area in the mining process. And it can return the corresponding access sequential patterns when the projection tree only has one branch. Secondly, in the recommendation stage, we use the sequential patterns mined by the improved FLWAP-mine algorithm to achieve personalized recommendation. Web access sequential patterns are stored via a type of tree structure and the page matching is based on the tree structure. Then, the page recommendation algorithm based on access sequential pattern is proposed. The variable active window technology and the concepts of page selection attention degree and page average visit degree are adopted in this algorithm. Combining page selection attention degree and page average visit degree with the confidence of the rules as page recommendation degree produces the recommendation results. In the end, two experiments are carried out. The experimental results show that the improved FLWAP-mine algorithm performs better than the previous one and the recommendation algorithm based on access sequential pattern can improve the recommendation precision.