一种基于模糊C均值聚类的协同过滤推荐算法

摘要

针对协同过滤推荐系统中数据过于稀疏导致推荐质量差的问题,提出一种有效的针对稀疏评分的最近邻选择方法,使用奇异值分解方法将用户评分分解为不同的特征,以预测用户对他没有评过分的项目的喜好,实验结果表明,推荐结果的精确性得到明显提高.该方法有效地避免数据稀疏性对推荐质量造成的影响,在一定的近邻数下,算法的推荐质量提高了一倍.%According to the poor recommendation quality because of data sparsity in collaborative filtering recommendation system,we present an efficient neighbor selection method.The matrix of samples should be processed by singular value decomposition to obtain several characteristics,which can apply to predict users preferences for those items,According to experimental results,the accuracy of recommendation has been improved obviously.The method can effectively avoid the influence of recommendation quality caused by data sparsity,and the quality of recommendation algorithm doubled in certain of number of the nearest neighbors.