基于时间权重的协同过滤算法在电子商务中的应用

发布时间：2018-03-09 12:24

本文选题：推荐系统　切入点：协同过滤　出处：《湘潭大学》2015年硕士论文　论文类型：学位论文

【摘要】：随着互联网技术和海量数据处理技术的发展,人们已经进入了一个新的时代——从以前的信息缺乏来到了信息过量的时代。这种转变也给人们带来了巨大的难题:一方面对用户来说,如何从海量信息中找到自己的需要的、感兴趣的信息成为了一个挑战;另一方面对于商家来说,如何推销自己的商品,吸引住用户的眼球,也是一个值得研究的问题。为了解决上述难题,推荐系统应运而生了,成了商家与用户之间的桥梁,其目标就是为用户和商家信息提供更便利、更快捷的服务,不仅是用户能够发现感兴趣的东西,也能使商家把自己的商品推销出去,展现给需要的用户。推荐系统的这种作用和意义,在电子商务网站中尤为重要。因此,研究推荐算法在电子商务中的应用,将非常有科研意义。论文针对基于项目(Item-based)的协同过滤算法没有考虑时间权重这一问题,提出了增添时间考虑因素的改进型Item协同推荐算法,设计了基于时间权重的协同过滤电子商务推荐系统模型,以购书网站为应用背景,设计一个购物网站推荐模型,并对其性能进行了验证。本论文首先重点研究并对比了基于用户(User-based)的协同过滤算法和基于项目(Item-based)的协同过滤算法。然后提出将时间因素加入到推荐系统中并作为时间权重参与计算,这样可以更接近用户最新的兴趣状态。对基于项目和评分的相似度计算公式进行了改进,采用时间加权相似度计算公式代替原有公式。并且,为减少时间开销,算法结合了数据挖掘的聚类方法,项目空间先经过聚类来降低计算时间和空间开销,增加计算效率。最后,将改进的“基于项目内容和评分的时间型加权协同过滤算法”进行仿真实验,实验结果表明,改进算法在平均绝对偏差和平均消耗时间上均优于原基于项目的协同过滤算法,验证了改进算法的高效性。最后进行了基于协同过滤的电子商务推荐系统模型设计。以图书商城购书网站作为推荐模型的实际应用背景,设计了购物网站个性化推荐模型,主要内容包括:推荐模型的体系结构设计、主要功能分析、推荐流程分析、各个功能模块的设计以及后台数据库设计等。
[Abstract]:With the development of Internet technology and massive data processing technology, People have entered a new era-from a lack of information to an era of information overload. This change has also brought people a huge problem: on the one hand, for users, how to find their own needs from the mass of information, Information of interest has become a challenge; on the other hand, how to promote their products and attract the attention of users is also a problem worth studying. In order to solve these problems, the recommendation system came into being. It has become a bridge between merchants and users, whose goal is to provide more convenient and faster services for users and businesses. It is not only that users can discover what they are interested in, but that they can also sell their products. The role and significance of recommendation system is especially important in e-commerce websites. Therefore, the application of recommendation algorithm in e-commerce is studied. Aiming at the problem that the time weight is not taken into account in the collaborative filtering algorithm based on item-based, this paper proposes an improved Item collaborative recommendation algorithm which adds time factors. A collaborative filtering E-commerce recommendation system model based on time weight is designed, and a shopping website recommendation model is designed based on the application background of book purchase website. The performance is verified. Firstly, the collaborative filtering algorithm based on user User-based and the collaborative filtering algorithm based on item-based are studied and compared. Then the time factor is added to the recommendation system and used as a recommendation system. Time weights participate in the calculation, In this way, we can get closer to the latest state of interest of the user. The similarity calculation formula based on item and score is improved, the time-weighted similarity calculation formula is used instead of the original formula, and, in order to reduce the time cost, The algorithm combines the clustering method of data mining, the project space is first clustered to reduce the computation time and space overhead, and increase the computing efficiency. The improved time-weighted collaborative filtering algorithm based on item content and score is simulated. The experimental results show that the improved algorithm is superior to the original item-based collaborative filtering algorithm in terms of average absolute deviation and average consuming time. Finally, the model of E-commerce recommendation system based on collaborative filtering is designed. The personalized recommendation model of shopping website is designed with the book-shopping website as the practical application background. The main contents include: the architecture design of the recommendation model, the main function analysis, the recommendation flow analysis, the design of each function module as well as the background database design and so on.
【学位授予单位】：湘潭大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.3;F724.6

【参考文献】