一种面向稀疏数据的比率相似度计算方法

A ratio similarity calculation method for sparse data

针对传统协同过滤方法存在数据稀疏问题,该文提出了一种面向稀疏数据的比率相似度计算方法,该方法在相似度计算过程中仅基于用户全部的显式评分数据,并且不依赖于共同评分项。用户的未评分项目通过相似度计算结果和最近邻的评分数据进行预测,并将预测评分较高的项目推荐给用户,实现个性化推荐。实验在两个公开的数据集上进行,结果表明,在数据稀疏的情况,该方法下仍然能够实现较高的推荐精度。

Aiming at the data sparsity problem in the traditional collaborative filtering method, a ratio similarity calculation method for sparse data is proposed, which is only based on the user′s explicit rating data in the similarity calculation process, and does not depend on co-rated items. The user′s unrated items are predicted by the similarity calculation results and the nearest neighbor′s rating data, and the items with the higher predicted ratings are recommended to the user to implement personalized recommendation. The experiment was carried out on two public datasets. The experimental results show that in the case of sparse data, the method can still maintain a high recommendation accuracy.