基于加权多融合偏好与结构相似度的协同过滤算法

Collaborative Filtering Algorithm Based on Weighted Multi-fusion Preference and Structure Similarity

当推荐系统中用户评分数据集是稀疏数据时,使用基于评分相似度或基于结构相似度的传统协同过滤算法会增加最近邻选取误差。针对这一不足,综合用户对于类别的偏好情况,提出基于融合评分及类别偏好相似度的协同过滤算法。为更准确地发现相似用户,考虑用户在评分结构上存在的相似性,进一步提出加权多融合偏好及结构相似度度量方法。实验结果表明,该算法可减少平均绝对误差,提高推荐质量。

If the rating data set of users in the recommendation system consists of sparse data,it is not ideal to use the traditional collaborative filtering algorithm based on rating similarity or structure similarity, because of the large error in nearest neighbor selection. Aiming at the shortcomings of traditional collaborative filtering algorithms, this paper proposes a collaborative filtering algorithm fusing rating and class preference similarity. In order to measure the similarity between users more accurately, the measurement method of weighted multi-fusion preference and structure similarity is the further improvement of the measurement method of fusing score and class preference similarity. Through the experimental comparison, it is showed that the proposed algorithm can reduce the Mean Absolute Error（MAE） and improve the quality of recommendation.