一种结合评分重合度的协同推荐算法

协同推荐是信息个性化服务中广泛应用的推荐算法，协同推荐算法以宿主系统所观测到的用户评分作为实现推荐的数据依据。用户评分矩阵的稀疏性问题对协同推荐的各工作过程可产生直接或间接的影响，导致推荐服务的准确性下降。通过对稀疏性问题影响推荐系统方式的分析发现，一般协同推荐方法的项目相似度计算只注重项目在评分数值上的相关性，而忽视了项目之间评分的重合度对提高推荐质量所起的重要作用。通过将评分重合度融入到相似度计算中，提出了一种结合评分重合度的改进协同推荐算法，并在稀疏评分环境下将其与已有协同推荐算法进行了对比实验与分析，实验结果验证了所提算法在提高预测准确性上的有效性。

Collaborative filtering approach combined with rating overlap

Collaborative filtering has been the most widely employed personalized approach in recommender systems, it produces recommendations based on the user ratings observed by the host system. The issue of sparsity about the rating matrix can directly or indirectly affect the accuracy of recommendations. By analyzing the impacting ways of sparsity, it was found that the existing collaborative filtering approaches have emphasized the correlation between the rating values of items in calculating the item-based similarity, whereas the positive effect of rating overlay on improving the recommendation accuracy was not taken into consideration. By integrating the rating overlay with the classical similarity, this paper proposed an improved collaborative filtering approach. In contrast with the classical approach, the experimental results exhibit the effectiveness in promoting the prediction accuracy in the context of rating sparsity.