基于遗忘函数和项目流行度的推荐算法研究

发布时间：2018-03-10 02:00

本文选题：推荐算法　切入点：协同过滤　出处：《北方民族大学》2017年硕士论文　论文类型：学位论文

【摘要】：由于互联网技术日新月异的快速发展,电子商务已逐渐成为我们生活中不可或缺的一部分。绝大多数电子商务平台为我们提供了琳琅满目的商品,同时也为消费者提供了很多便利的服务。但是如何为消费者提供更加准确的推荐是当前各大电子商务平台急需要解决的问题,因此研究新的推荐技术或改良以前提出的方法也成为了时下的一个热门课题,而推荐技术中最关键的就是推荐算法的研究。在目前存在众多的推荐算法中,协同过滤推荐算法使用的最多。尽管协同过滤推荐应用如此之广,但是随着注册用户和商品数量增加,协同过滤算法的准确性和推荐效果变差,有效的解决此问题可以提高用户的忠诚度和满意度。基于以上研究背景,本文的研究主要针对推荐算法中的协同推荐算法存在的相似度计算不准确的问题,提出了一种基于遗忘函数和项目流行度的改进算法,主要思路就是对艾宾浩斯遗忘函数和项目流行度因素进行改进。以传统的用户相似度算法为基础,通过融合以上两个条件得到用户综合的相似度算法,为用户推荐其感兴趣的产品。最后借助Movielens数据集,对本文提出的推荐方案进行检验,并与传统的推荐算法进行比较,实验结果表明本文提出的推荐算法有效的提高了推荐精度。为了更好的检验本文的推荐效果,本文开发了一个推荐算法的WEB应用,以更好的提高用户体验效果。
[Abstract]:With the rapid development of Internet technology, e-commerce has gradually become an indispensable part of our life. At the same time, it also provides a lot of convenient services for consumers. However, how to provide more accurate recommendation to consumers is an urgent problem that each major e-commerce platform needs to solve. Therefore, the research on the new recommendation technology or the improvement of the previous methods has become a hot topic, and the most important part of the recommendation technology is the research of the recommendation algorithm. At present, there are many recommendation algorithms. Although collaborative filtering recommendation is widely used, as the number of registered users and commodities increases, the accuracy and recommendation effect of collaborative filtering algorithm become worse. Solving this problem effectively can improve the loyalty and satisfaction of users. Based on the above research background, this paper mainly focuses on the problem of inaccurate similarity calculation in the collaborative recommendation algorithm. In this paper, an improved algorithm based on forgetting function and item popularity is proposed. The main idea is to improve the Ibinhaus forgetting function and project popularity factors, which is based on the traditional user similarity algorithm. Through the fusion of the above two conditions, the similarity algorithm of user synthesis is obtained to recommend the products of interest to the user. Finally, with the help of Movielens data set, the proposed recommendation scheme is tested and compared with the traditional recommendation algorithm. Experimental results show that the proposed recommendation algorithm effectively improves the recommendation accuracy. In order to better test the recommendation effect, this paper develops a recommendation algorithm WEB application to improve the user experience.
【学位授予单位】：北方民族大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.3

【参考文献】