基于耦合相似度的矩阵分解推荐方法

发布时间：2018-05-09 18:49

本文选题：推荐系统 + 相似度　；参考：《计算机科学》2016年04期

【摘要】：随着因特网和信息技术的高速发展,信息过载现象越来越严重。推荐系统能够给个人和商家(例如电子商务和零售商)提供个性化的推荐。数据稀疏性和分数预测质量问题被公认为是现存推荐系统中的主要挑战。当前绝大多数推荐系统技术都依赖于协同过滤方法,它主要利用用户-项目评分矩阵来表示用户和项目之间的关系。一些研究利用附加信息来提高推荐准确性,但是,绝大多数现存的引入项目之间关系的方法并不能很好地用于预测和推荐,因为其假设项目属性之间是独立同分布的,而实际上项目(或用户)的属性之间是存在耦合关系的。由此提出了基于属性耦合关系的矩阵分解模型,它能有效地刻画项目之间的耦合相关性,从而更加合理地预测用户对项目的评分。实验结果表明,所提出的模型在热启动和冷启动的推荐准确性方面均优于传统的推荐算法。
[Abstract]:With the rapid development of Internet and information technology, the phenomenon of information overload is becoming more and more serious. Recommendation systems can provide personalized recommendations to individuals and businesses (e. G. E-commerce and retailers). The problem of data sparsity and score prediction quality is recognized as the main challenge in the existing recommendation systems. At present, most recommendation systems rely on collaborative filtering methods, which mainly use the user-item scoring matrix to express the relationship between users and items. Some studies use additional information to improve recommendation accuracy, but the vast majority of existing methods to introduce relationships between items are not well used to predict and recommend because they assume that the attributes of the items are independently and equally distributed. In fact, there is a coupling between the attributes of the project (or user). A matrix decomposition model based on attribute coupling relationship is proposed, which can effectively describe the coupling correlation between items, thus more reasonably predicting the users' scores on the items. The experimental results show that the proposed model is superior to the traditional recommendation algorithm in the accuracy of recommendation for hot start and cold start.
【作者单位】：辽宁工程技术大学电子与信息工程学院;
【基金】：国家青年科学基金(61003162) 辽宁省高等学校杰出青年学者成长计划(LJQ201303038)资助
【分类号】：TP391.3

【相似文献】