基于时间因素影响的矩阵分解算法的研究

发布时间：2018-01-30 15:51

本文关键词： 推荐系统时间因素矩阵分解回归分析　出处：《内蒙古大学》2017年硕士论文　论文类型：学位论文

【摘要】：现代网络技术水平的提高带来了信息过载的问题,用户很难从巨大的信息中找到用户需要的内容。为了解决这个问题,推荐系统被提出并延伸出了诸多改进。这其中就包括缓解了数据稀疏性问题的矩阵分解模型以及诸多改进模型等。它们能从用户的行为数据中挖掘出用户的行为规律,从而给用户更准确地推荐结果。然而,矩阵分解技术也存在着一些问题。其中一个问题便是它只关注用户的行为数据,并没有从时间的角度来考虑用户的兴趣变化。虽然有通过融入时间因素改进矩阵分解的算法,但是它们所用的时间规律并没有从真实的用户行为数据中提取。本文从给矩阵分解模型增加时间因素出发。首先,调查统计了大量的真实用户行为数据,得出了用户兴趣度和物品流行度在时间上所具有的规律。然后对统计所得的结果进行回归分析,通过分析时间因素以怎样的规律影响用户兴趣和物品流行度,进而拟合出相应的曲线。然后将物品流行度的时间曲线和用户兴趣度的时间曲线分别与基于偏置的矩阵分解模型相融合,并分析在时间因素影响下用户和物品偏置对推荐结果的影响。最后,本文通过实验验证了基于时间因素改进矩阵分解模型的准确性确实有所提高。
[Abstract]:The improvement of modern network technology has brought about the problem of information overload. It is very difficult for users to find the content that users need from the huge information. In order to solve this problem. Recommendation systems are proposed and extended to many improvements, including matrix decomposition models that alleviate the problem of data sparsity and improved models. They can extract user behavior from user behavior data. Rules. However, matrix decomposition technology also has some problems. One of the problems is that it only pays attention to user behavior data. The change of interest of users is not considered from the point of view of time, although there is an algorithm to improve matrix decomposition by incorporating time factors. However, the time rules they use are not extracted from the real user behavior data. In this paper, we add time factors to the matrix decomposition model. First, we investigate a large number of real user behavior data. The time law of user interest and article popularity is obtained. Then the results of statistics are analyzed by regression analysis, through the analysis of time factors to affect user interest and article popularity. Then fitting the corresponding curve, and then combining the time curve of article popularity and the time curve of user interest with the matrix decomposition model based on bias. Finally, this paper proves that the accuracy of the improved matrix decomposition model based on time factors has been improved.
【学位授予单位】：内蒙古大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.3

【相似文献】