基于相似度质量的混合协同过滤算法

Hybrid Collaborative Filtering Algorithm Based on Quality of Similarity

在传统协同过滤算法中一直面临着冷启动和数据稀疏等问题,导致推荐信息不够准确。通过分析基于用户的协同过滤算法和基于物品的协同过滤算法的各自特点提出一种新的混合协同过滤算法。改进相似度的计算方式来提高相似度的精准度,从近邻相似度的均值和标准差出发对两种协同过滤算法进行加权结合,同时引入控制因子提高预测的精度。以Movie Lens数据集进行实验验证,以平均绝对误差作为实验的测试标准。实验结果表明,在评分矩阵极度稀疏的条件下该算法提高了推荐的准确度。

In the traditional collaborative filtering algorithm has been facing a cold start and data sparseness and other issues,resulting in the recommendation information is not accurate enough. A new hybrid collaborative filtering algorithm is proposed by an.alyzing the characteristics of user-based collaborative filtering algorithm and item-based collaborative filtering algorithm. This pa.per combines the weighted mean of two similar filtering algorithms with the mean and standard deviation of the similarity,and intro.duces the control factor to improve the precision of the prediction. Experiments are carried out with the Movie Lens dataset,and theaverage absolute error is used to measure the results. The experimental results show that the proposed algorithm improves the accura.cy of the proposed algorithm when the scoring matrix is extremely sparse.