改进用户评分相似度的协同过滤推荐算法

Collaborative Filtering Recommendation Algorithm Based on Improved User Similarity

在协同过滤推荐算法中,传统的相似度计算方法在计算时未能考虑用户共同评分项目数量差异、评分数值差异、项目热门度差异和用户兴趣随时间因素变化差异的问题,导致相似度计算结果不准确,推荐结果准确率较低。针对这些问题,提出一种改进用户评分相似度的协同过滤推荐算法。通过在余弦相似度与修正余弦相似度的基础上引入评分数值差异与项目热门度差异的修正因子,来缓解用户评分差异与项目权重的影响;其次根据用户的兴趣随时间因素变化的特点,提出时间衰减因子,以捕捉用户的兴趣偏好的动态变化;最后引入权重因子将改进的两种相似度计算方法相结合,从而缓解用户共同评分项目数量差异所导致的问题,提升推荐结果的准确性与现实意义。通过使用MovieLens数据集进行对比实验,相比传统基于用户的协同过滤算法MAE值平均降低了5.43%,证明提出的改进算法能有效提高推荐结果的准确性。

In the collaborative filtering recommendation algorithm,the traditional similarity calculation method fails to take into account the differences in the number of items jointly rated by users,the differences in scoring values,the differences in item popularity,and the differences in user interests over time,resulting in similarity calculation results.Inaccurate,the recommenda⁃tion result has a low accuracy rate.Aiming at these problems,this paper proposes a collaborative filtering recommendation algo⁃rithm to improve the similarity of user ratings.On the basis of cosine similarity and modified cosine similarity,a correction factor for the difference in score value and the difference in item popularity is introduced to alleviate the influence of user score difference and item weight.Time decay factor to capture the dynamic changes of users’interests and preferences;finally,a weight factor is in⁃troduced to combine the two improved similarity calculation methods,so as to alleviate the problem caused by the difference in the number of items jointly rated by users,and improve the accuracy of the recommendation results.realistic meaning.Compared with the traditional user-based collaborative filtering algorithm,the MAE value is reduced by 5.43%on average by using the MovieLens dataset,which proves that the proposed improved algorithm can effectively improve the accuracy of the recommendation results.