结合置信度和SVD的协同过滤算法

Collaborative Filtering Algorithms Based on Confidence Level and SVD

为了解决基于传统模型的协同过滤算法的数据稀疏性与冷启动问题,引入置信度参数,并结合隐式反馈信息,提出了两种基于奇异值分解(SVD)的协同过滤算法,CSVD和NCSVD。CSVD算法在基于偏置的矩阵分解模型上引入了置信度参数,以改进模型偏置项没有针对物品规模根据每个评分调整偏置权重的问题,NCSVD在此基础上引入隐式反馈信息,改善了冷启动问题,在真实数据集上的实验证明表明,其能有效提高SVD系列算法的推荐精度。

In this paper, two collaborative filtering models, namely CSVD and NCSVD, are investigated to deal with two problems of the traditional model-based collaborative filtering algorithms, in particular, the problem of data sparsity and the problem of cold start. In the CSVD model, a confidence factor is introduced to the matrix factorization model to adjust the bias weight of each item according to its size. The NCSVD model then solves the cold start problem by using an implicity feedback factor based on the CSVD model. Experimental results on realistic datasets shows that our proposed models have better prediction results than the state of the art methods.