一种新的基于用户的协作过滤推荐算法

协作过滤是应用最为广泛的推荐技术，通常提供预测评分作为推荐。提出一种新的协作过滤算法，采用概率形式．即预测用户喜欢商品的概率来推荐。算法采用基于用户的思路，扩展最近邻算法，通过训练建立预测值和概率形式之间的映射模型，考察相似用户的评价提供概率形式的推荐。实验结果表明该算法能够提供比较准确的预测。

A New User-based Collaborative Filtering Recommendation Algorithm

Collaborative filtering, which is widely used recommendation algorithm, usually provides predicted ratings as recommendation. A new algorithm is proposed, which uses a probability value as the output showing the chance that a user might like an item. It employs a user-based approach. At first, a mapping model between the predictions and the probability values is constructed by training. The recommendations are then generated based on the opinions of similar users. Experimental results show that the algorithm can give accurate predictions.