基于信息匹配度的混合推荐算法

Hybrid recommendation algorithm based on matching degree of information

针对推荐系统中协同过滤算法的冷启动问题无法解决、矩阵分解算法推荐解释性较差、单一推荐算法适应性不强的问题,提出了一种基于信息匹配度的协同过滤和矩阵分解的动态权重混合算法(UIBCF-MF)。该算法考虑了各算法推荐结果和历史信息的不匹配度,将基于用户的协同过滤算法和基于物品的协同过滤算法与传统的矩阵分解算法按照特定规则进行混合,提出了一种基于皮尔逊相关系数的主观评分规范化方法。当新用户或新物品出现时,对推荐结果做二次矩阵分解,解决了冷启动问题并具有一定的可解释性,真实MovieLens电影数据集上的实验结果表明,该算法在稀疏矩阵中能有效降低推荐误差,提高推荐精度。

Collaborative filtering(CF)algorithm has cold-start problem which is impossible to solve,matrix factorization algorithm recommendation is poor in interpretability,and single recommendation algorithm is not very adaptable.To solve the above problems,this paper proposes a dynamic weight hybrid algorithm(UIBCF-MF)based on information matching degree of collaborative filtering and matrix decomposition.UIBCF-MF takes into consideration the mismatch between recommendation results and historical information of each algorithm,and combines user-based collaborative filtering,item-based collaborative filtering algorithm and traditional matrix decomposition algorithm according to specific rules.This paper proposes the subjective score normalization method based on Pearson correlation coefficient.By calculating the predictive score,when new users or items appear,the proposed algorithm makes a quadratic matrix decomposition to the recommendation results.The algorithm solves cold-start problem and the interpretability is also improved.The experimental results on the real MovieLens movie dataset show that the proposed algorithm can effectively reduce the recommendation error and improve the recommendation accuracy in the sparse matrix.