Research And Improvement Of Collaborative Filtering Algorithm Of Similarity And The Scalability Problem

In recent years, with the gradual increase of the electronic commerce market speed amplification and network shopping turnover, huge commodity information with the life of the consumer, how to conveniently provide personalized information for the users, the recommendation technology emerge as the times require.This paper has made a thorough research on e-commerce recommendation system, a detailed analysis of the current situation and Prospect of technology was recommended in the field of electronic commerce of various personalized recommendation. Collaborative filtering recommendation algorithm as the current research hotspot, its inherent sparsity, scalability and recommendation quality is the focus of the study.Aiming at these problems, this paper presents the similarity model based on improved MSD, improved MSD and combined with the similarity score of the inherent context information, than the traditional PCC and MSD measurement in the recommendation accuracy, the best prediction and worst prediction performance has the promotion; and the traditional inverted index technology is introduced into collaborative filtering(CF) algorithm in the improved fuzzy clustering, and jumping inverted collaborative filtering algorithm based on index structure, which is better than conventional CF algorithm in the nearest neighbor user generated time significantly reduced and the precision is better than CF algorithm and the traditional.