Abstract:

Collaborative filtering techniques have been widely used in various recommender systems. However, they are likely to suffer from the data sparsity problem which often results in inaccurate recommendation results. In view of this problem, this paper proposes a collaborative filtering recommendation algorithm based on Sigmoid function, by analyzing users' invocation records of services. Firstly, the relationship between a user and the invocation number of service is analyzed to acquire the user's interest in contents of service by using TF-IDF algorithm. Then, a Sigmoid function is used to calculate the user's interest in function of service according to the invocation number. Finally, personalized service recommendation is performed by combining the user's contents interestingness with the function interestingness. The experimental results show that this algorithm can effectively alleviate the data sparsity problem and thus achieve better prediction accuracy.

Key words: collaborative filtering, Sigmoid function, data sparsity, recommender system, users?? interest

摘要：

协同过滤推荐技术被广泛用于各个推荐系统，但它仍然存在着用户评分数据稀疏性问题，可能导致推荐结果不准确。针对该问题，提出了一种采用Sigmoid函数的协同过滤推荐算法。首先，分析用户兴趣与其调用服务的次数之间的关系，利用TF-IDF算法计算用户对服务内容的兴趣度；其次，定义一个Sigmoid函数，根据服务调用次数计算用户对服务功能的兴趣度；最后，基于内容兴趣度和功能兴趣度计算用户兴趣相似度完成协同过滤算法，实现个性化的服务推荐。实验证明，该方法能有效缓解数据稀疏性问题，提高了推荐质量。

关键词: 协同过滤, Sigmoid函数, 数据稀疏性, 推荐系统, 用户兴趣