论文部分内容阅读
推荐系统帮助用户过滤无用信息并预测其可能感兴趣的产品。在推荐系统中,协同过滤是应用最为广泛的方法之一。然而,传统的协同过滤方法是在产品维度上计算用户相似度,而且在计算相似度时无法考虑邻居用户的影响。因此,该类方法往往受到高维度、数据稀疏等问题的困扰。为此,本文提出一种基于用户兴趣传播的协同过滤方法,在兴趣维度上计算用户相似度,同时考虑了兴趣在不同用户间的传播。该方法不仅可以有效防止冷启动和数据稀疏问题,而且具有较高的预测准确度。在标准数据集MovieLens上的测试结果表明了本文算法的有效性。
The recommendation system helps users filter out unwanted information and predict what products they might be interested in. In the recommendation system, collaborative filtering is one of the most widely used methods. However, the traditional collaborative filtering method is to calculate user similarity in the product dimension, and the influence of neighbor users can not be considered when calculating the similarity. Therefore, such methods are often plagued by problems of high dimension and data sparseness. To this end, this paper proposes a collaborative filtering method based on user interest propagation, which calculates the user similarity in the interest dimension while considering the spread of interest among different users. This method not only can effectively prevent cold start and data sparseness, but also has higher prediction accuracy. The test results on the standard dataset MovieLens demonstrate the effectiveness of the proposed algorithm.