基于协同过滤技术的电子商务个性化系统研究

【作者】 王亮；

【导师】 王世卿；

【作者基本信息】 华东师范大学 ， 软件工程， 2008， 硕士

【摘要】 随着网络的普及和信息技术的日新月异,电子商务系统在为用户提供越来越多选择的同时,其结构也变得更加复杂,用户经常会迷失在大量的商品信息空间中,无法顺利找到自己需要的商品。电子商务个性化系统在这种情况下应运而生。电子商务个性化系统在电子商务平台上根据用户的资料与行为向用户推荐商品,帮助用户找到所需商品,从而顺利完成购买过程。其应用前景广泛,受到国内外。用户的广泛关注。协同过滤技术是目前电子商务个性化系统中应用最早和最为成功的技术之一,它的基本思想:为用户找到他真正感兴趣的内容的好方法是首先找与他兴趣相似的用户,然后将这些用户感兴趣的内容推荐给此用户。一般采用最近邻技术,计算用户之间的距离,然后利用目标用户的最近邻居用户对商品评价的加权评价值来预测目标用户对特定商品的喜好程度,系统从而根据这一喜好程度来对目标用户进行推荐。协同过滤最大优点是对推荐对象没有特殊的要求,能处理非结构化的复杂对象,如音乐、电影。论文研究了当前个性化推荐的主流技术——协同过滤技术对该算法中影响推荐质量的稀疏性问题和影响用户满意度的推荐完整性问题进行了深入分析,引入了组合推荐技术对协同过滤算法进行改进,设计出了能够实现论文所提出的推荐策略的实验仿真。论文对所研究的改进算法进行了仿真实验,经过实验验证,改进算法在推荐的准确性、完整性、多样性等方面均优于传统算法,特别是在稀疏的用户评价数据集上体现出了良好的推荐性能。论文设计了一个电子商务个性化系统框架,完成了通用流程,为现实的电子商务个性化系统提供了有益的参考。更多还原

【Abstract】 With the popularity of network and information technology with each passing day, E-commerce system to provide users with more and more choices at the same time, its structure has become more complex, users often get lost in a large number of goods in the information space, could not find their own need for commodities. Personalized e-commerce system came into being in such a case. Personalized E-commerce systems in E-commerce platform based on the user’s information to the user behavior and recommend products to help users find the goods in order to the successful completion of the purchase process. Its wide range of applications have attracted users at home and abroad by the attention.Collaborative filtering technology is one of the earliest and most successful application of the technology in personalized E-commerce system. Its basic idea is: it found a good way for users to find the contents of his genuine interest is to find him and his interest in similar users first. Then these users interested in the content of recommended to the user. The general nearest neighbor, using the user’s historical preferences information calculated the distance between the user and then use the nearest neighbor target users for goods evaluation of the weighted evaluation of target users to predict the extent of the preferences of specific commodities, so as to the basis of this system preference to the recommended target users. Collaborative filtering biggest advantage is recommended objects do not have a special request, be able to handle unstructured complex objects, such as music, movies.This paper focus on the current recommendation of the mainstream of personalized technology - Collaborative filtering technology impact of the algorithm in the light of the recommendation of quality issues and customer satisfaction impact on the recommendation of the integrity of the in-depth analysis of the issue, recommended the introduction of a combination of collaborative filtering algorithms, technology to improve the design of the paper can be achieved by the recommended strategy for the simulation. Research papers on the algorithm to improve the simulation experiments, the experimental verification, recommended improved algorithm in the accuracy, integrity, diversity and so on is better than the traditional method, especially in light of the evaluation of user data sets reflect on the recommendation of a good performance. We designed a personalized E-commerce systems framework, the completion of a common process for the reality of personalized e-commerce system provides a useful reference.更多还原