一种基于Sigmoid函数的改进协同过滤推荐算法

随着电子商务和社交网络的蓬勃发展, 推荐系统逐渐成为数据挖掘领域的重要研究方向。推荐系统能够从海量信息中定位用户兴趣点, 提供个性化服务。协同过滤算法能够有效分析用户偏好, 提供合适的推荐服务。针对评分矩阵稀疏时传统协同过滤算法性能很差的问题, 提出一种基于Sigmoid函数的改进推荐系统算法。利用Sigmoid函数对不同项目进行建模, 得到项目的平均受欢迎程度; 利用Sigmoid函数对不同用户进行建模, 将评分映射为用户对项目的喜好程度; 根据用户对项目喜好程度应该与项目平均受欢迎程度贴近的原则进行评分预测。在两组真实数据集合上的实验结果表明, 该算法较好地解决了数据稀疏性问题, 能够有效提高传统算法的预测准确性。

Improved collaborative filtering recommender algorithmbased on Sigmoid function

As the rapid development of the electronic commerce and social network, recommender systems have become one of the most important research areas in data mining field. Recommender systems can identify users' interest out of humorous information in order to provide personalized service. Collaborative filtering (CF) is efficient in extracting users' preferences and making proper recommendations. To address the data sparsity problem of classic CF algorithms and improve the performance, this paper introduced an improved algorithm based on Sigmoid function. Different items were modeled with Sigmoid function in order to capture their popularity, while different users were modeled to map ratings into preferences. Predictions were made according to that preferences should keep consistent with popularities. Experimental results on two real world datasets show the proposed method can alleviate the sparsity problem and are effective to improve the performance of classic CF algorithms.