融合用户相似度与评分信息的协同过滤算法

Collaborative filtering algorithm based on user similarity and rating information

推荐系统利用机器学习的技术进行信息过滤,准确地定位用户需要的信息,并且能够预测用户对目标项目的喜好程度。但是由于新用户和新项目的存在,传统的协同过滤推荐系统面临着冷启动问题的挑战。为了解决协同过滤推荐系统中用户冷启动问题,设计了融合用户相似度与评分信息的协同过滤算法(SR-CF)。该算法用基于人口统计学的推荐算法找出用户基本信息之间的相似度,再根据最速下降法对用户评分矩阵进行更新,从而产生对目标用户的推荐。基于Moive Lens公开数据集的实验结果表明,所设计的算法在保证推荐准确率的同时提高了推荐的覆盖率,能有效解决用户冷启动问题。

Recommendation systems by using machine learning techniques to filter information can accurately locate user information, and predict whether a user would like a given resource. As traditional col- laborative filtering systems have to deal with the cold-start problems caused by new users and items, a collaborative filtering algorithm based on user similarity and rating information（SR-CF） is designed to solve the new-user cold-start problem. The algorithm can calculate similarities between users＇ basic information by the recommendation algorithm based on demography, and then update the matrix of ratings information with the steepest descent method to generate recommendations for users. The experimental results on MoiveLens dataset demonstrate that SR-CF can improve the accuracy of the recommendations and has higher coverage. Thus it can effectively deal with the cold-start problems caused by new-users.