Abstract: Traditional collaborative filtering computes the similarity between users based on the rating value, however, the data sparseness causes the inaccurate of similarity computing. Aiming at this problem, this paper proposes a novel method based on quantity of rating information. It translates the amount of user who rates item with the same value to quantity of rating information, and combines the item-rating to calculate the similarity. The experimental results show that the proposed can relief the effect of data sparseness and improve the accuracy of predicted rating comparing to the traditional algorithm.

Key words: similarity, amount of rating user, quantity of rating information, collaborative filtering

摘要： 传统的协同过滤算法中，依靠用户评分大小计算用户间相似度，但是评分数据稀疏性使相似度计算不够准确。针对此问题，提出了基于评分信息量的相似度计算方法；在推荐系统中项目有多种可选评分，该方法将参与评分的用户数量转换为评分信息量，以此结合用户评分大小计算相似度。实验结果表明，相对于传统协同过滤算法，该方法在一定程度上减少了评分数据稀疏性带来的负面影响，有效地提高了预测评分准确性。

关键词: 相似度, 评分用户数量, 评分信息量, 协同过滤