基于SVD与模糊聚类的协同过滤推荐算法

Collaborative Filtering Recommendation Algorithm Based on SVD and Fuzzy Clustering

协同过滤为个性化推荐解决信息过载问题提供了方案,然而也存在着数据的稀疏性、可扩展性等影响推荐质量的关键问题.我们提出了一种基于奇异值分解(SVD)与模糊聚类的协同过滤推荐算法,通过引用物理学上狭义相对论中能量守恒的方法以保留总体特征值的数目,较为准确地确定降维维度,实现对原始数据的降维及其数据填充.另外,再运用模糊聚类的方法将相似用户进行聚类,从而达到减少邻居用户搜索范围的目的.在Movie Lens与2013年百度电影推荐系统比赛等不同数据集上的实验结果表明,该算法能够提高推荐质量.

Collaborative filtering provides a solution for the personalized recommendation to solve the problem of information overload. But the problems of data sparsity and scalability are the serious factors affecting the recommendation quality. To solve these problems, we propose a collaborative filtering algorithm based on singular value decomposition and fuzzy clustering. We retain the number of the total characteristic value through the theory of energy conservation in the special relativity in physics, so as to determine the dimension of dimension reduction. In addition, by using the fuzzy clustering, we also reduce the search range of the neighbors. Compared with traditional collaborative filtering recommendation algorithm in the different data sets of Movie Lens and 2013 Baidu movie recommendation system, the proposed algorithm performs better in the recommendation quality.