基于最大期望和协同过滤算法的研究与应用

Research and Application of Algorithm Based on Maximum Expectation and Collaborative Filtering

推荐系统中新用户在信息搜索中易出现信息稀疏的问题,以致给用户推荐相关模块的时候带来了极大困难。针对该问题,采用人口统计学中的最大期望算法对用户进行聚类找到近邻用户,然后将其作为协同过滤算法的输入。由于用户对不同项目的评分表明他们需求,相同用户评价的项目中存在一定的需求关联性。而且随着个人需求的变化,这种关联度也逐渐在变化。所以通过引入一个时间权重函数的形式,给出一种基于用户需求变化的协同过滤算法,缓解传统协同过滤推荐算法的短板。可以追踪到用户的需求,进而预测评分矩阵。通过实验和比较,该算法有助于解决用户的评分矩阵稀疏性问题,从而提高推荐质量。

The problem of sparse information is easily found in the information search for new user in recommendation system, and the difficulty is produced when recommending the relevant module for users. In view of the problem, the maximum expectation algorithm in demographics is adopted to cluster users for neighboring users, and then it is regarded as input of collaborative filtering algorithm. AS the user＇ s scores on different projects show that they demand,in the evaluation of the project from same user there is certain demand rele- vance. And this kind of relevance degree is gradually changing with the change of individual demand. Therefore, a cooperative filtering al- gorithm based on the change of user demand is put forward by introducing a time weight function, which alleviates the shormess of tradi- tional cooperative filtering recommendation algorithm. Can track the needs of users,and then predict the score matrix. According to exper- iments comparison, this algorithm can help solve the problem of sparseness of user＇ s scoring matrix and the recommendation quality is improved.