The Research Of E-Commerce Recommendation System Based On Association Rules And User Preference

The rapid development of e-commerce has brought an explosive growth of the number of users and goods. As a way to help users to filter information and provide a personalized service, recommendation system was applied in the mass data processing soon. The core of the system is recommendation algorithms, association rules algorithm is one of the most popular algorithms. Among the same algorithms, FP-Growth algorithm is the most famous. However, in current recommendation systems which based on FP-Growth algorithm still have some problems: lack of confidence degree, the transaction without discrimination, lower coverage, vulnerable models, less personalized, etc. So the FP-Growth algorithm needs to be improved to meet the current needs of the e-commerce recommendation system.In this thesis, FP-Growth algorithm is improved in many strategies, and based on the optimized algorithms, the thesis design and implement an e-commerce recommendation system. Specific tasks include:based on support and confidence degree parallel to implement the algorithm on a strong association rules; the design of aging degree and interest degree to overcome the shortcomings affairs without discrimination; the TOP-K algorithm based on hierarchical and user preference to solve the empty recommendation of "Long Tail commodity"; In addition, design dual threshold mechanism and blacklist against attacks, etc. The recommendation system is designed and implemented in java technology on Hadoop platform. Experimental results show that through correlation analysis on a large number of actual data, the recommendation system based on improved FP-Growth algorithm improves the accuracy rate, the recall rate and predicted coverage rate significantly.