一种基于用户兴趣联合相似度的协同过滤算法

A collaborative filtering algorithm based on user interest and joint similarity

在推荐系统中数据稀疏性和推荐时效性是经常面对的问题,为了更好地反映不同用户在不同阶段的邻域相关性,从而能够挖掘出评分项目中所隐含的个性化信息,在基于用户的协同过滤算法预测评分过程中将联合相似度与用户兴趣的时序信息相结合,首先融合覆盖评分信息的用户间的协同相似度、偏好相似度和轨迹相似度等3种相似度,通过参数调节不同度量的权重及相似度阈值形成联合相似度以获取用户有效的邻居数目;其次在联合相似度计算过程中引入反映时间权重的Logistic函数以提高推荐的时效性;最后进行实验,结果表明,所提出的方法与经典算法相比,不仅提高了精度,而且可以更有效地预测用户的真实评分。

In recommender system,sparsity and timeliness are the usually faced problems.A new multiplied similarity computation with time-information was proposed,which applied to user-based collaborative filtering method,to better reflect the neighbor dependency of different users at different stages,then to dig out the personalized information that implied in the rating items.Firstly,the fusion of joint similarity method,collaborative similarity,preference similarity and trajectory similarity were addressed to gain the users neighbor numbers by adjust the relative weight and the threshold of similarity.Then,the time-weight Logistic function in the calculation process of joint similarity was used to improve the timeliness.Finally,the experiments were carried out.The results showed that the proposed method had better recommendation accuracy than the classical algorithms,and could more effectively predict the user’s true rating.