基于云填充和混合相似性的协同过滤推荐算法的研究

摘要

针对传统推荐算法的相似性度量准确性不高及数据极端稀疏性等问题，提出一种基于云填充和混合相似性的协同过滤推荐算法。首先通过云模型填充用户-项目评分矩阵，然后对相似性度量方法进行改进，将基于时间序列的用户间影响力融合到基于Jaccard 系数的相似性度量方法中。在 MovieLens 数据集上的验证结果表明，改进后的算法提高了推荐精度同时在一定程度上克服了数据稀疏性的影响。%A collaborative filtering recommendation algorithm based on cloud model filling and hybrid similarity was pro-posed to measure the similarity of the traditional recommendation algorithm with low accuracy and extreme sparsity of data. First,the user-item rating matrix was filled by the cloud model,and then the similarity measure method was improved,and the influence of the user based on time series was fused to the similarity measure method based on the Jaccard coefficient. The validation results on MovieLens data sets show that the improved algorithm can improve the recommendation accuracy and overcome the influence of data sparsity to a certain extent.