Abstract: The traditional collaborative filtering technology based on rating can not find the right criteria. It is difficult to find exact relation for large ratings aggregate of items, so the expression of personalize is influenced. A collaborative filtering recommendation algorithm based on multiple ranked choosing domains is proposed, which computes comparison of preference value by relevance of sliding matched items in choosing domains. It proposes a novel conservative similarity algorithm for profile matrix, and implements predicted ratings by the similar users. Experimental results show that the proposed method can efficiently ease the sparsity of data, and significantly improve recommender precision in predicted ratings.

Key words: collaborative filtering recommendation, multiple ranked choosing domains, conservative similarity algorithm, predicted ratings

摘要： 传统的基于用户评分的协同过滤推荐系统无法找到合适的评分标准，对大量的评分数据挖掘不足，影响了用户的个性化表达。针对该问题，提出一种基于多序选择域的协同过滤推荐算法，采用选择域滑动匹配寻找项目关联性算法计算偏爱比较值，通过相似特征矩阵进行未评价项目的预测评价。实验结果表明，该推荐算法通过预测未评价项目可有效缓解数据的稀疏性，提高了推荐质量。

关键词: 协同过滤推荐, 多序选择域, 相似度保守算法, 预测评价

CLC Number: 

• TP391