Distributed collaborative filtering recommendation model based on two-phase similarity

The recommendation system based on collaborative filtering is one of the most popular recommendation mechanisms. However, with the continuous expansion of the system, several problems that traditional Collaborative Filtering recommendation algorithm (CF) faced such as cold startup, accuracy, and scalability are worsened. In order to address these issues, a Distributed Collaborative Filtering recommendation model based on Two-Phase similarity (DCF-TP) is proposed. DCF-TP is based on Weighted Distance Similarity Measure (WDSM), a new measure created in this paper. According to WDSM, the similarity of the users is calculated and the similarity matrix of users is obtained, meanwhile, in line with the co-occurrence matrix method, the similarity of items is counted, getting the co-occurrence matrix of the items. With regard to the similarity matrix of users, their preferences are endowed with weights and the new preferences matrix of users is received. Besides, on the basis of the co-occurrence matrix of the items and the new preferences matrix of users, the nearest neighbor item is found and a more accurate recommendation to the target user is given. Furthermore, in terms of the parallel computing framework, the distributed implementation of DCF-TP is completed. All these experiments are done on MovieLens dataset. The results show that DCF-TP overcomes the problem of cold startup and has a qualitative leap both in the aspects of precision and recall ratio. With the increasing numbers of the computing nodes, the distributed algorithm has achieved higher linear speedup.