基于Jaccard相似度和位置行为的协同过滤推荐算法

发布时间：2018-05-31 09:22

本文选题：Jaccard相似度 + 位置行为　；参考：《计算机科学》2016年12期

【摘要】：协同过滤是现今推荐系统中应用最为成功且最广泛的推荐方法之一,其中概率矩阵分解算法作为一类重要的协同过滤方式,能够通过学习低维的近似矩阵进行推荐。然而,传统的协同过滤推荐算法在推荐过程中只利用用户-项目评分信息,忽略了用户(项目)间的潜在影响力,影响了推荐精度。针对上述问题,首先利用Jaccard相似度对用户(项目)做预处理,而后通过用户(项目)间的位置信息挖掘出其间的潜在影响力,成功找到最近邻居集合;最后将该邻居集合融合到基于概率矩阵分解的协同过滤推荐算法中。实验证明该算法较传统的协同过滤推荐算法能够更有效地预测用户的实际评分,提高了推荐效果。
[Abstract]:Collaborative filtering is one of the most successful and widely used recommendation methods in today's recommendation systems. As an important collaborative filtering method, probability matrix decomposition algorithm can be recommended by learning low-dimensional approximate matrix. However, the traditional collaborative filtering recommendation algorithm only uses user-item scoring information in the process of recommendation, neglecting the potential influence between users (projects) and affecting the recommendation accuracy. To solve the above problems, Jaccard similarity is used to preprocess the user (project) firstly, and then the potential influence is mined through the location information between users (project), and the nearest neighbor set is found successfully. Finally, the neighbor set is fused into the collaborative filtering recommendation algorithm based on probability matrix decomposition. Experimental results show that the proposed algorithm is more effective than the traditional collaborative filtering recommendation algorithm in predicting users' actual score and improving the recommendation effect.
【作者单位】：南京工业大学计算机科学与技术学院;中国人民解放军73677部队;
【基金】：国家自然科学基金(61203072) 江苏省重点研发计划(社会发展)(BE2015697)资助
【分类号】：TP391.3

【相似文献】