基于粗糙集的马田系统研究及其在银行直接营销客户分类中的应用

发布时间：2018-07-12 16:59

本文选题：粗糙集 + 马田系统　；参考：《南京理工大学》2015年硕士论文

【摘要】：外资银行凭借先进的营销策略对我国的商业银行产生了巨大冲击,国内客户对产品和服务的需求趋于多样化和个性化,众多依靠传统模式开拓市场的国内商业银行无法保持良好的市场竞争力,国内商业银行亟需改变传统营销策略。直接营销作为一种以客户需求为中心的新型营销模式已在国外得到广泛应用,国内银行采用直接营销模式将有助于提升市场竞争力,而直接营销模式得以有效实施的关键在于找到合适的方法通过客户分类定位目标客户。常用分类方法通常需要对数据分布进行假设,而马田系统作为一种新的模式识别方法是基于数据进行分析完成分类的,而且它能删除冗余变量,提取有效信息,真正实现系统降维,马田系统已成功应用于多个行业的分类问题中。本文针对马田系统在筛选特征变量方面的不足引入粗糙集理论,构建基于粗糙集的改进马田系统方法,并将改进的马田系统应用于银行直接营销客户分类问题中,定位目标客户。本文的主要研究内容包括以下两个方面：(1)基于粗糙集的马田系统理论研究传统马田系统结合正交表和信噪比筛选有效特征变量以优化基准空间,但有学者研究表明,正交表与信噪比方法在筛选有效特征变量方面存在不足。本文引入粗糙集理论替代正交表和信噪比对特征变量进行选择,构建基于粗糙集的马田系统分类方法,以更好地优化基准空间、改善分类效果。(2)基于粗糙集的马田系统应用研究为研究国内银行开展直接营销的客户分类问题,选取UCI数据集中某葡萄牙银行直接营销活动相关数据作为分析数据,进行基于粗糙集的马田系统应用研究。分别用基于粗糙集的马田系统和传统马田系统分析银行客户数据进行客户分类,并对比二者筛选的有效特征变量个数及分类准确率。结论表明：与传统马田系统相比,基于粗糙集的马田系统不仅提升了分类准确率,而且减少了有效特征变量个数,可以进行准确分类并简化信息收集工作。基于粗糙集的马田系统方法可以应用于银行直接营销客户问题中进行准确的客户定位,有助于国内银行有效实施直接营销模式。与此同时,基于粗糙集的马田系统方法可以应用于其他分类问题中。
[Abstract]:With the advanced marketing strategy, the foreign banks have a great impact on our commercial banks. The domestic customers' demand for products and services tends to diversify and individualize. Many domestic commercial banks relying on the traditional mode can not maintain a good market competitiveness. The commercial banks in the country need to change the traditional marketing strategy. As a new marketing mode centered on customer demand, marketing is widely used abroad. The direct marketing mode of domestic banks will help to improve the market competitiveness. The key to the effective implementation of the direct marketing model is to find the right way to locate the target customers through customer classification. The data distribution is usually supposed to be assumed, and the Martin system is a new pattern recognition method based on data analysis, and it can delete redundant variables, extract effective information, and truly realize the system reduction. The Martin system has been successfully applied to the classification of many industries. This paper is aimed at Martin system in this paper. The shortage of feature variables is introduced into the rough set theory, and the improved Martin system method based on rough sets is constructed, and the improved Martin system is applied to the customer classification problem of direct marketing of banks. The main research contents of this paper include the following two aspects: (1) the research of Martin system theory based on Rough Set Traditional Martin system combines orthogonal tables and signal-to-noise ratio to filter effective feature variables to optimize the reference space, but some scholars have shown that there is a shortage of orthogonal tables and signal-to-noise ratio methods in screening effective feature variables. The field system classification method is used to better optimize the reference space and improve the classification effect. (2) the application of Martin system based on rough sets is used to study the customer classification problem of direct marketing in domestic banks. The data of the direct marketing activities of a Portuguese bank in a UCI data set are selected as the analysis data, and the Martin system based on rough sets is carried out. Use the Martin system based on rough set and the traditional Martin system to analyze the customer data of the bank, and compare the number of effective feature variables and the classification accuracy of the two parties. The conclusion shows that compared with the traditional Martin system, the Martin system based on rough set not only improves the classification accuracy, but also reduces the classification accuracy. Without the number of effective feature variables, we can classify and simplify the information collection. The Martin system method based on rough sets can be applied to the accurate customer location in the bank direct marketing customer problems and help the domestic banks to implement the direct marketing model effectively. At the same time, the rough set based Martin system method can be used. It is applied to other classification problems.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：F832.33

【参考文献】