基于评分可信度的协同过滤融合方法

Collaborative filtering fusion method based on rate credibility

为了解决基于项目和基于用户的推荐结果的融合问题,提出了基于评分可信度的协同过滤融合方法。该方法在推荐对象集合中计算评分数大于平均评分数的对象作为信任子群,在信任子群上计算能够使推荐的平均绝对误差最小的融合参数λ。由计算得到的最优融合参数λ对基于项目和基于用户的推荐结果进行融合,作出推荐。实验结果表明,该算法有效提高了过滤推荐的精准度和可靠性,具有良好的推荐效果。

To overcome several limitations in the research of collaborative filtering（ CF） fusion,this paper presented a CF recommendation fusion algorithm based on rate credibility. This method calculated trustworthy subset which rated times above the average rated times from both user-based and item-based recommendation set,calculated optimal λ on trustworthy subset repeatedly. The method used optimal λ to fuse user-based and item-based recommendation set. Experimental results show that this algorithm can effectively achieve higher recommendation accuracy and reliability and better recommendation results.