信用卡信息记录平台构建及聚类方法研究与应用

发布时间：2018-01-26 23:02

本文关键词： 信息记录平台 B/S架构 J2EE框架聚类方法划分方法　出处：《吉林大学》2017年硕士论文　论文类型：学位论文

【摘要】：生活在21世纪的我们,已经处在一个高度发达的信息科学技术时代,信息科学技术的发展快速无比,尤其在经济、政治、文化、军事等方面的应用更是与日俱增,信息技术在使得人类社会文明取得了很大的进步与发展的同时,在互联网金融行业的应用也逐渐广泛。为了高效地管理人们信用卡的消费信息,提高信息处理效率,信用卡信息记录平台已逐渐地应用到企业对信用卡信息的管理当中。但在信息技术广泛应用的同时,数据库中存储的信息量也快速增长,使得人们高效提取有效信息的难度逐步增加,由此导致对数据挖掘技术应用的迫切需求,将数据挖掘算法应用在信息量较大的平台上也是非常有必要的。近些年来,数据挖掘已经逐渐被应用到各个行业领域当中,例如互联网金融、车载网、电子地图导航应用等各个领域。作为当前的前沿科研课题与很有发展前景的技术,数据挖掘不仅涉及到数据库技术、机器学习等多种技术,还与统计学、概率理论等多种基础理论紧密相关。聚类分析算法是数据挖掘方法之一,近年来人们在这方面的研究探索取得了很大的进步,不同的聚类分析方法也纷纷出现,提出了各种基于不同思想的聚类方法,并加以验证,给出实现,最后应用在合适的场景之中。在涉及到人工智能科学技术的各个领域当中,这些方法都有所涉及,在相应的领域当中也取得了众多成果。然而,当下的互联网金融行业之中,有一些问题需要解决,更需合适的聚类方法应用于其中。本文工作主要是构建信用卡信息记录平台,选择、改进聚类方法,最后将其应用于平台。具体内容主要包括:(1)构建了信用卡信息记录平台。基于B/S架构与Java平台,针对目标需求,进行开发,设计了对应的功能模块,分别为安全管理、信息采集、信息管理、统计报表、信息聚类和平台管理等模块,支持根据需求开发新的功能模块。前端使用HTML、CSS、Java Script设计页面,后端使用Java语言编写,并采用Struts2、Hibernate、Spring等框架的相关技术,数据库选择在互联网上应用比较广泛并且开源的My SQL。(2)深入研究了基于划分的聚类算法。划分的方法主要有基于k平均的和k中心的方法,在设计的信息平台上,信用分是作为衡量消费者信用的主要方式,从上限100递减,每个分数段作为一个信用等级划分,到下限60为止。虽然k中心算法所划分的数据结果明显,由于其随机选择初始中心点,聚类结果具有不可控性,每一组聚类结果分布与每一个信用等级范围不是很接近。而本文通过对传统的k中心算法进行改进,选取信息记录条数最多的k组数据作为中心点,同时通过对迭代次数的限制,然后进行聚类划分,可以提高接近程度来提高管理消费者信用等级的效率,从而高效地观察不同信用等级消费者的行为习惯,制定相应的信用等价折扣策略,提高商业利益。本文最后对文中的工作进行了总结,对算法的优缺点给出了分析,并对未来的工作做出了展望。
[Abstract]:We have been living in the twenty-first Century, in a highly developed information technology era, the rapid development of information science and technology incomparable, especially in the economic, political, cultural, military and other aspects of the application of information technology is in progress and grow with each passing day, the development of the civilization of human society made great and application in Internet banking the industry is gradually widely. In order to efficiently manage the credit card consumer information, improve the efficiency of information processing, credit card information recording platform has been gradually applied to the enterprise credit card information management. But in the extensive application of information technology at the same time, the amount of information stored in the database is growing rapidly, which makes people efficient extraction the effective information is more difficult, which led to an urgent need for application of data mining, data mining algorithm application platform in large amount of information Is also very necessary. In recent years, data mining has been applied to various fields, such as Internet banking, mobile network, the electronic map navigation application. As a frontier research topic of current and promising technology, data mining is not only related to the database technology, machine learning and other technologies also, with statistics, probability theory and other basic theory are closely related. Clustering algorithm is one of the methods of data mining in recent years, people explore the research in this area has made great progress, with the method of clustering analysis have also appeared, put forward a variety of different clustering methods based on the idea, and verified, and presents the finally, the appropriate application in the scene. In artificial intelligence involves various fields of science and technology, these methods have been involved in the field, when Also made many achievements. However, in the current Internet financial industry, some problems need to be solved, more suitable to the application of clustering method. The main work of this paper is to construct the credit card information recording platform selection, improved clustering method, and finally applied to the platform. The specific contents include: (1) the construction of the credit card information recording platform. The architecture of B/S and Java based platform, development target, demand, function module design of the corresponding, respectively for the safety management, information collection, information management, statistical statements, information clustering and platform management module, support the development of new functional modules using HTML front-end according to requirements. CSS, Java, Script page design, is the use of Java language, and the use of Struts2, Hibernate, Spring and other related technical framework, database is widely used on the Internet and open source My S QL. (2) studied clustering algorithm based on the division. The division methods are mainly K and K based on the average center, in the design of the information platform, credit is as the main measure of consumer credit, decrease from the upper limit of 100, each fraction as a credit rating, to the lower limit 60 so far. Although the k center algorithm divided data the results are obvious, because of its random selection of initial centers, the clustering result is not controllable, the distribution of each group clustering and each credit rating range is not very close. And through the center of the traditional K algorithm is improved, the number of records selected information K group most data as the center point, at the same time through the iteration limit, then clustering can improve the proximity to improve the efficiency of the management of consumer credit rating, to effectively observe different credit rating Finally, we summarize the work in this paper, analyze the advantages and disadvantages of the algorithm, and make prospects for future work.

【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.13

【参考文献】