188宝金博页面版

http://www.paper.edu.cn - 1 - 中国中国科技论文在线科技论文在线协同过滤推荐系统研究 牟丹，刘辉*（辽宁工程技术大学经济管理学院，辽宁 阜新 123000） 摘要： 推荐系统作为电子商务网站的一种重要的营销工具， 使消费者在面对海量的商品信息时得到有价值的购买建议的同时， 也提高了网站的销售业绩。 随着网络与计算机技术的进步，电子商务推荐系统得到广泛的研究与发展，服务越来越智能化、个性化。电子商务下的个性化信息推荐系统在理论和实践上都得到了很大发展， 其核心部分就是对推荐方法的研究， 选择哪种推荐方法对于推荐系统的效果和效率至关重要。 协同过滤算法是个性化推荐系统中常用的方法， 由于用户对项目的评价部分不多， 使得数据稀疏性对推荐质量和实时性影响较大。 本文对电子商务环境下的个性化信息推荐服务进行了深入地研究， 并针对个性化推荐方法中的协同过滤算法进行了详尽地分析，给出了协同过滤算法所面临的挑战，在此基础上，对基于聚类的协同过滤推荐算法进行了改进。该算法在用户-项目评分数据库的基础上，首先离线计算项目之间的相似度，结果保存在数据库中；其次对每个项目聚类结果进行用户聚类，计算用户之间的相似度， 找到用户的最近邻居集合； 最后根据用户最近邻居集合预测对每个项目的评分，产生推荐给目标用户。本文改进了算法的可行性和有效性，从而可以有效解决用户评分数据极端稀疏情况下传统方法存在的不足， 有效提高了电子商务推荐算法的推荐质量。 关键词：个性化信息推荐；相似性；协同过滤；电子商务推荐系统 中图分类号：TP315 The Recommendation System Research Of on the Collaborative Filtering Algorithm Mu Dan, Liu Hui (COLLEGE ECONOMIC MANAGEMENT,LIAONING TECHNICAL UNIVERSITY, LiaoNing FuXin 123000) Abstract: As an important marketing tool for the E-commerce web sites, the recommended system provides customers with valuable advices from millions of item information, so as to help promote the marketing performance of the web sites. With the progress of the Internet and the computer technology, the E-commerce recommended system receives deeper research and development which make it produce more intelligent and more personal services. On the aspect of theory research and factual application, E-business personalized recommended systems have developed rapidly. Recommendation approaches are the core part of recommended system, so which approach is adopted is crucial to the success of recommending quality as well as efficiency. Collaborative filtering algorithm is the prevalent recommendation method at present. However, because the proportion of user rating data is few, the sparsity of data influences the quality and the real-time requirement of recommendation. A collaborative filtering recommendation method based on items and customers clustering was proposed in this paper to solve these problems: This paper studied deeply E-commerce recommendation system, analyzed the collaborative filtering algorithm and challenges which collaborative filtering recommendation approach suffered from, then we presented an improved collaborative filtering algorithm based on clustering. The followings are the main steps of improved algorithm: Firstly, to compute offline items similarity based on the Users-Items rating database, and to save results in database; Secondly, for every clustered sets , to compute users similarity and find out the user's neighbors sets. Finally, to predict every item's rating according to the nearest neighbors and to generate recommendation to target users.We tested the improved algorithm through programming, the experiment result proved that this algorithm is feasible and effective. Compare to traditional collaborative filtering algorithm, this algorithm can overcome the sparsity of user's rating information and generate better recommendation results. Keywords:Personalized information recommendation; similarity; collaborative filtering; E-business recommendation system 作者简介：牟丹，（1978-），女，讲师，计算机应用及信息管理. E-mail: fxmudan_cn@sina.com