基于影响集与修正权重的协作过滤推荐方法

协作过滤推荐算法是构造推荐系统最成功的推荐技术之一。提出了一种基于影响集与修正权重的协作过滤方法:该方法选择被用户共同评分的项目集计算项目间相似性,过滤其相似度超过预定阈值的项目集作为影响集,然后设置项目共同出现的频次参数进行权重调整;并结合影响集与权重调整作出评分预测。实验结果说明了该方法的可行性与有效性。

Collaborative Filtering Recommendation Algorithm Based on Influence Sets and Weight Adjusting

Collaborative filtering recommendation algorithm is one of the most successful technologies for building recommender systems. A collaborative filtering algorithm based on Influence Sets and weight adjusting is proposed in this paper. First, common user ratings over any two items are selected to compute the similarity between these two items, and only those items whose similarity value exceed presetting threshold are selected as the set of supporting items. Then a parameter of the number that two items appears together is used to adjusted similarity weight, and prediction ratings are produced by combining Influence Sets and weight adjusting. Finally, the Experimental result shows that this proposed algorithm is feasible and effective in practice.