基于用户兴趣的协同过滤算法研究

发布时间：2018-07-05 20:14

本文选题：用户兴趣 + 时间窗　；参考：《安徽理工大学》2017年硕士论文

【摘要】：随着web 2.0时代的到来,数据量呈指数式增长。面对海量数据人们无法从中迅速找到自己所需资源。为了解决人们在资源检索和选择上的问题,学者们提出了推荐系统。在推荐系统中,协同过滤算法是如今应用最广泛的推荐算法之一。但在实践应用上,传统协同过滤算法在推荐时未考虑到针对用户兴趣变化及项目自身属性进行推荐;从而影响了推荐的质量。为了解决此类问题,本文主要针对用户兴趣及项目属性这两方面展开了着重研究,提出相关改进与创新如下:1)研究分析现有推荐算法的现状与不足;并通过用户已评价项目的属性及评分构建出用户兴趣模型;有效解决了用户兴趣不易捕捉的问题。2)针对传统协同过滤算法在推荐中忽视了推荐项目自身属性;因此提出了基于用户兴趣的协同过滤算法,该算法结合了项目自身属性和用户兴趣模型给出了用户兴趣度,促使推荐过程中忽视的项目属性问题得以解决。3)考虑到用户兴趣随时间变化产生的偏移,引入了艾宾浩斯遗忘定律及滑动时间窗来体现出用户的兴趣偏移,并对用户兴趣协同过滤算法进行了改进优化。实验数据选用于经典的MovieLens数据集;对提出的用户兴趣协同过滤算法及改进后算法进行了实验。验证结果表明,我们提出的算法在推荐中有效解决了用户兴趣捕捉及项目冷启动问题,推荐的质量也得到提升。
[Abstract]:With the arrival of the web 2.0 era, the amount of data increases exponentially. In the face of massive data, people can not quickly find their own resources. In order to solve the problem of resource retrieval and selection, scholars put forward the recommendation system. Collaborative filtering is one of the most widely used recommendation algorithms in recommendation systems. In practice, however, the traditional collaborative filtering algorithm does not take into account the change of user interest and the properties of the item itself, which affects the quality of recommendation. In order to solve this kind of problem, this paper mainly focuses on the user interest and the item attribute, and puts forward the related improvement and innovation as follows: 1) Research and analysis the current situation and the insufficiency of the existing recommendation algorithm; The user interest model is constructed through the attributes and scores of the items evaluated by the user, and the problem of user interest capture is solved effectively. 2) in view of the traditional collaborative filtering algorithm, the attributes of the recommendation items are ignored in the recommendation process. Therefore, a collaborative filtering algorithm based on user interest is proposed, which combines the properties of the project itself and the user interest model to give the user interest degree. In order to solve the problem of item attribute, which is neglected in the recommendation process, this paper introduces the Ibinhaus' law of forgetting and sliding time window to reflect the deviation of user's interest, considering the deviation of user's interest over time. The collaborative filtering algorithm of user interest is improved and optimized. The experimental data are selected from the classical Movie Lens dataset, and the proposed collaborative filtering algorithm and the improved algorithm are tested. The verification results show that the proposed algorithm can effectively solve the problems of user interest capture and project cold start, and improve the quality of recommendation.
【学位授予单位】：安徽理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.3

【参考文献】