【作者】 李炜；

【导师】 周水庚；

【作者基本信息】 复旦大学 ， 计算机技术， 2013， 硕士

【摘要】 随着信息技术和互联网的高速发展,电子商务也随之迅速发展,人类逐渐从信息匮乏的时代渐渐走向了大数据的时代,信息繁琐和找不到人们需要的信息成为现在电子商务发展的瓶颈。在这个信息技术的新时代,无论对于信息的创造者还是信息的阅读者,都被淹没在大量的选择之中,信息创造者需要定向投递广告,信息阅读者需要找到符合自身愿望的产品,推荐系统的出现就是为了解决这一矛盾的重要工具。推荐系统,顾名思义就是为电子商务的用户来推荐其感兴趣的产品。它根据用户个人的属性,个人的喜欢,购买习惯来向其推荐相应的产品。本文的推荐是基于保险的,目前全世界对于保险的推荐还处于一个空白的阶段,如果只是套用现有的推荐算法,当用户属性不是很明确或者购买浏览行为不够多的情况下,比较难以对其进行个性化的推荐,推荐的效果就完全失去了意义。本文首先调研了国内外现有主流电子商务网站的推荐系统,在此基础上分析了X1保险电子商务网站的现状,对开发保险个性化推荐系统进行了比较详尽的需求分析,然后设计出了一个保险个性化推荐系统的系统框架,该系统框架包括推荐系统的实施层、推荐系统的引擎层、推荐系统的数据／知识层三个模块。其次,在推荐系统的详细设计阶段,量化了客户属性和产品属性,根据各个推荐算法的需要,设计出了数据库的模式,包括推荐算法需要的原始数据表和推荐结果表,重点设计了适合于各个场景和各个客户类型的个性化和非个性化的推荐算法,包括基于统计的推荐算法、基于关联规则的推荐算法、基于用户协同过滤的推荐算法、基于条目的协同过滤推荐算法和基于内容的推荐算法。然后为了验证推荐算法的正确性,对各个推荐算法进行了测试。最后,设计出了一个系统部署技术方案,将推荐系统和X保险电子商务网站进行了对接。更多还原

【Abstract】 With the rapid development of the information technology and network, e-commerce is booming. People step into the big data times from the era of information scarcity gradually. Information is so much and the complexity of searching information have become the bottleneck of the development of the e-commerce. In the new times of the information technology, both producers and readers, are overwhelmed by so many choices. Producers need to deliver targeted advertisements, while readers need to find products that meet their expectation. However, recommendation system is the important tool to deal with this contradiction.Recommendation system, by definition is used to recommend some specific products to specific users in e-commerce. It is a software system which is based on the customer’s personal attributes, personal likes, buying habits to recommend appropriate products to specific customers. However, this paper is based on insurance recommendation which is an empty page in the world. If we use original algorithms, in the circumstance of lack of users’attributes and buying or browsing behavior, it will be very difficult to recommend appropriate products to users and the target of recommending is lost.Firstly, we investigate the existing domestic and international main e-commerce recommendation system. Based on this, we analysis the status of the X e-commerce insurance website and requirement of the development of the personalized recommendation system. And then we devised a system framework, which includes implementation layer, the engine layer and the data/knowledge layer. Secondly, in the detailed design stage of recommendation system, we quantify the customer attributes and product attributes. According to each recommendation algorithm, we designed the schema of the database, which includes the raw data tables and the result table. Importantly, we designed four recommendation algorithms that is suitable all kinds of scenes and customers. These algorithms include the statistical recommendation algorithm, the association rules recommendation algorithm, the user-based collaborative filtering recommendation algorithm, the item-based collaborative filtering recommendation algorithm, the content-based recommendation algorithm. And then in order to verify the correctness of the algorithm, we tested these algorithms. Finally, we devised a system deployment technology solution to interchange with the X insurance e-commerce website.更多还原