【摘要】 针对传统协同过滤算法的冷启动、推荐精度低等问题,提出基于用户属性和项目属性的协同过滤算法以及它们两者的融合推荐算法。在计算用户相似度时,提出用户年龄、性别和职业属性差异度,并与皮尔逊相关系数加权结合;在计算项目相似度时,提出项目类型标签和项目被评分时间,并将两者与项目余弦相似度融合。最后将上述两种算法的推荐结果进行加权融合。实验结果表明,改进的融合推荐算法相比其他4种算法在平均绝对误差率(mean absolute error,MAE)和时间性能方面有更好的推荐结果,并且能够在有新用户和新项目出现的情况下明显提高推荐系统的推荐质量。更多还原

【Abstract】 Traditional collaborative filtering algorithm has the problems of cold start and low recommendation accuracy.A collaborative filtering algorithm based on user attributes,a collaborative filtering algorithm based on item attributes and fusion recommendation algorithm of the former two,were proposed.When calculating user similarity,the user’s age,gender and occupational attribute differences were proposed and weighted with Pearson correlation coefficient.When calculating item similarity,the item type labels and item scoring time were proposed,and the two were combined with the project cosine.Finally,the recommendation results of the above two algorithms were weighted and fused.The experimental results showed that the improved fusion recommendation algorithm had better recommendation results in terms of mean absolute error(MAE) and time performance than the other four algorithms,and could be used in situations where new users and new items appeared.This significantly improved the recommendation quality of the recommendation system.更多还原