基于隐语义模型和聚类算法的电子商务个性化推荐系统的研究与实现

发布时间：2018-04-24 01:08

本文选题：电子商务 + 协同过滤　；参考：《北京邮电大学》2017年硕士论文

【摘要】：随着互联网和电子商务的高速发展,购物变得越来越方便。一方面网络上琳琅满目的商品满足了人们的各种需求,另一方面随着商品的增多,找到自己中意的商品也越来越难,从而很容易使得顾客产生疲劳直至失去购物兴趣,这就是“信息超载”。电子商务个性化推荐系统作为解决信息超载问题的一种方案,它让电子商务网站适应每一个消费者并为其提供具有个性化的商品展现平台和购物体验。本文主要研究电子商务个性化推荐系统。首先分析了个性化推荐系统在国内外的研究成果,并对常用的推荐技术进行分析。之后提出一种基于隐语义模型和聚类算法的协同过滤算法(CF-LFMC),并基于该方法设计了一种基于移动端的个性化推荐系统。本文所做的主要工作有以下三个方面:1.分析传统基于项目的协同过滤算法(Item-based Collaborative Filtering,IBCF)中存在的数据稀疏性、时效性和冷启动问题,使用隐语义模型和聚类算法,并引入时间函数和项目初始属性填充的方法对传统IBCF算法进行改善,提出一种基于隐语义模型和聚类算法的协同过滤算法(CF-LFMC)。2.实验测试验证了提出的CF-LFMC算法,结果表明算法在准确率上优于传统算法,达到了预期的效果。3.设计了一种基于CF-LFMC算法的电子商务个性化推荐系统。本文描述了系统的架构,然后从客户端到服务器的角度描述了系统的设计,并对其中的模块进行了详细的分析与实现。
[Abstract]:With the rapid development of internet and e-commerce, shopping becomes more and more convenient. On the one hand, there are a wide variety of goods on the Internet to meet people's needs. On the other hand, with the increase of commodities, it is becoming more and more difficult to find the goods that they like, which makes it easy for customers to become tired and lose their interest in shopping. This is information overload. E-commerce personalized recommendation system as a solution to the problem of information overload, it allows e-commerce websites to adapt to every consumer and provide them with personalized commodity display platform and shopping experience. This paper mainly studies the e-commerce personalized recommendation system. Firstly, the research results of personalized recommendation system at home and abroad are analyzed, and the commonly used recommendation technology is analyzed. Then a collaborative filtering algorithm based on hidden semantic model and clustering algorithm is proposed. Based on this method, a personalized recommendation system based on mobile terminal is designed. The main work of this paper is as follows: 1. This paper analyzes the problems of data sparsity, timeliness and cold start in Item-based Collaborative filtering algorithm, and uses hidden semantic model and clustering algorithm. The traditional IBCF algorithm is improved by introducing time function and item initial attribute filling method, and a collaborative filtering algorithm based on implicit semantic model and clustering algorithm is proposed. The experimental results show that the proposed CF-LFMC algorithm is superior to the traditional algorithm in accuracy and achieves the desired effect of .3. An e-commerce personalized recommendation system based on CF-LFMC algorithm is designed. This paper describes the architecture of the system, then describes the design of the system from the point of view of client to server, and analyzes and implements the modules in detail.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.3

【参考文献】