基于用户数据挖掘的博库图书商城推荐方法研究

发布时间：2018-03-08 11:18

本文选题：博库图书商城　切入点：用户数据挖掘　出处：《哈尔滨理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着电子商务领域数据量的迅速增长,电子商务已进入以数据为基础的时代,电商都希望抓住这一契机,将数据有效利用,个性化推荐已成为电商数据应用的重要切入点,其作用就是为用户提供高质量的个性化推荐服务。通过收集大量用户数据并挖掘隐含的用户偏好,向用户推荐个性化商品,刺激用户的购物欲望,从而提高销售额,获得更大的利润。除此之外,个性化推荐还能挖掘出用户潜在的购物意向,向客户推荐到目前为止未接触过但以后可能购买的商品信息。不仅能为用户带来优质的服务,还能为电商企业带来前所未有的利润,这也是个性化推荐广泛应用的原因之一。因此,电子商务企业的个性化推荐不仅可以提高企业的服务质量,还可以促进用户购买。本文以博库图书商城为研究对象,分析商城产生数据的类型,特点及用户数据应用中存在的问题,进而分析商城数据挖掘过程及推荐方法,将用户数据分为文本评论数据,基本特征数据,交易行为数据。在评论的数据挖掘中,将虚假评论过滤,基于情感词提取图书属性,构建图书属性库,将用户—图书评分转化为用户—属性评分;在用户基本特征和交易行为特征的挖掘中,挖掘衍生属性,利用FCM聚类方法将用户分为高价值用户,中等价值用户,普通价值用户,低价值用户,在四个类内利用改进的用户相似性方法计算用户相似性,利用用户相似性填充用户—属性评分稀疏矩阵;基于填充后的矩阵得出目标用户偏好的最近邻居集,计算用户对未评属性的评分,将目标用户对商品的所有属性评分均值进行排序,形成该用户的Top-N推荐列表。完成推荐,通过实验证明推荐方法设计的准确性。通过对博库图书商城的用户数据挖掘及推荐方法的设计,一方面充分利用了商城产生的用户数据,结合协同过滤推荐理论,丰富数据挖掘理论及其应用;另一方面通过为博库图书商城设计个性化推荐方法,达到精准化地向用户推荐图书的目的,对网上图书类商城的经营提供帮助。
[Abstract]:With the rapid growth of data volume in the field of electronic commerce, electronic commerce has entered a data-based era. E-commerce hopes to seize this opportunity, make effective use of data, and personalized recommendation has become an important entry point for the application of e-commerce data. Its function is to provide high quality personalized recommendation service for users. By collecting a large amount of user data and mining hidden user preferences to recommend personalized products to users, stimulate the user's desire to buy, thereby increasing the sales volume. Besides, personalized recommendation can also discover the potential shopping intention of the user, and recommend to the customer information that has not been contacted so far but may be purchased in the future. It can not only bring good service to the user, It can also bring unprecedented profits to e-commerce enterprises, which is one of the reasons why personalized recommendation is widely used. Therefore, the personalized recommendation of e-commerce enterprises can not only improve the service quality of enterprises, This paper analyzes the types and characteristics of the data generated by the mall and the problems existing in the application of user data, and then analyzes the process of data mining and the recommended method of the shopping mall. The user data is divided into text comment data, basic characteristic data and transaction behavior data. In the data mining of comments, the false comments are filtered, the book attributes are extracted based on emotional words, and the library of book attributes is constructed. The user-book score is transformed into the user-attribute score, and in the mining of the user's basic characteristics and transaction behavior characteristics, the derived attributes are mined, and the users are divided into high value users, middle value users and ordinary value users by using FCM clustering method. In the four classes, the improved user similarity method is used to calculate the user similarity, the user similarity is used to fill the user-attribute score sparse matrix, and the nearest neighbor set of the target user preference is obtained based on the filled matrix. Calculate the user's rating of the unevaluated attributes, rank the target user's average value of all the attributes of the item, form the user's Top-N recommendation list, and complete the recommendation. Through experiments, the accuracy of the design of recommendation method is proved. Through the design of user data mining and recommendation method of Boku Book Market, on the one hand, it makes full use of the user data generated by the mall, and combines with the theory of collaborative filtering recommendation. On the other hand, by designing the personalized recommendation method for the book mall of Boku, we can accurately recommend the books to the users, and provide help for the management of the online book mall.
【学位授予单位】：哈尔滨理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：G239.23;F724.6;F274

【相似文献】