BP-CF hybrid recommendation algorithm: Implementation and analysis

作者：Yu Ruiguo; Zhang Minjie; Wang Zan^*; Yuan Zhenxia

来源：ICIC Express Letters, Part B: Applications , 2015, 6(1): 227-232.

• BP neural networks
• Hybrid algorithms
• Hybrid recommendation
• K-nearest neighbors
• Personalized recommendation
• Recommendation algorithms
• Sparsity
• Sparsity problems

Recommendation algorithm makes personalized recommendation by applying knowledge discovery. Among all recommendation algorithms, the k-nearest neighbor collaborative filtering (CF) is the most widely used. However, the sparsity problem makes the accuracy hardly to improve. In this paper we implement BP neural networks (simplified as BP)-CF hybrid algorithm to use the significant part of the rating matrix maximumly. By modelling with the relatively dense part of rating matrix using BP neural networks, we reduce the MAE on MovieLens dataset from 0.77 to 0.68.

• 出版日期2015
• 单位天津大学