基于标签权重评分的推荐模型及算法研究

发布时间：2018-10-19 08:36

【摘要】：推荐系统已经被越来越频繁地应用到电子商务网站与一些社交网站,在提高用户满意度的同时也带来了巨大的商业利益.然而,当前的推荐算法由于原始数据的不完整性以及算法本身处理数据的特殊性,导致推荐效果不理想.例如,某些推荐系统会产生冷启动、复杂兴趣推荐困难、解释性差等问题.为此,该文提出一种基于标签权重评分的推荐系统模型(Label-Weight Rating based Recommendation,LWR),旨在使用一种较为简洁的方式——标签权重评分来获取用户最准确的评价和需求,并通过改进当前的一些推荐算法来处理标签权重评分数据,从而生成对用户的推荐,最后以标签权重评分的形式向用户展示推荐结果并作出合理的解释.扩展实验中,通过电影推荐实验,证明了该文技术的有效性和可行性.
[Abstract]:Recommendation system has been used more and more frequently in e-commerce websites and some social networks, which not only improves user satisfaction, but also brings huge business benefits. However, due to the incompleteness of the original data and the particularity of the algorithm itself, the current recommendation algorithm does not have ideal recommendation effect. For example, some recommendation systems may have some problems such as cold start, difficult recommendation of complex interest, poor interpretation and so on. In this paper, a recommendation system model based on tag weight scoring (Label-Weight Rating based Recommendation,LWR) is proposed, which aims to obtain the most accurate evaluation and demand of users by using a simple method, that is, tag weight scoring. Some current recommendation algorithms are improved to deal with the tag weight scoring data, so as to generate recommendations to the user. Finally, the recommended results are presented to the user in the form of tag weight score and reasonable explanation is made. In the extended experiment, the effectiveness and feasibility of this technique are proved by the film recommendation experiment.
【作者单位】：哈尔滨工业大学计算机科学与技术学院;
【基金】：国家自然科学基金(61472099,61003046) 国家“九七三”重点基础研究发展规划项目基金(2012CB316200) 国家科技支撑计划(2015BAH10F00)资助~~
【分类号】：TP391

【参考文献】