Research And Application On The Retail Bank’s Customer Accurate Classification Model Based On Hadoop

As a new era of reform progress in financial markets and the innovative application of Internet technology,our country commercial bank present at a prosperous situation,but also facing unprecedented challenges and tests.In the current open and diversified competitive landscape,how to constantly improve the competitiveness of Banks is the focus of the banking industry.As an important part of the banking business,retail banking is developing rapidly and becoming an important source of bank profit.At present,China’s residents’ investment management awareness is gradually awakening,and the processing demand of personal assets tends to be diversified,which brings new opportunities for the development of retail banking business.In the face of the demands of diversified clients,it is one of the important means for Banks to attract customers by subdividing customers and understanding their different needs and providing differentiated services to customers.In order to establish a more accurate classification model for retail bank customers,this paper mainly carries on the following work.This paper makes an in-depth study of the current retail banking customer classification method and proposes a two-dimensional classification rule based on both customer value and business type.The current classification of retail bank customers mainly focuses on customer value or customer loyalty single dimension,which can only predict the value or loyalty of a customer but can not predict which type of business the customer is more interested in.This paper presents a kind of two-dimensional classification rule based on customer value and business type.In the customer value dimension,the customer is divided into four categories: target potential,gold card,platinum card and private bank.According to the characteristics of the retail banking business,customers are divided into three categories: partial asset class,partial liability class,and partial financial management.According to this classification rule,there are 12 types of retail banking customers,so that the bank can further understand the customer’s interest based on the customer’s value.In this paper,the advantages and disadvantages of a C4.5 algorithm are analyzed,and a C4.5 improvement algorithm based on rough set theory and CAIM criterion is proposed.The improved algorithm is used to deal with the continuous attribute with a data discretization algorithm based on the CAIM criteria,which effectively reduces the information loss of C4.5 algorithm in the processing of continuous attributes.The improved C4.5 algorithm USES attribute reduction algorithm based on rough set theory to make attribute reduction before constructing a decision tree,which excludes redundant attributes that are not related to classification in the decision table.To verify the effectiveness of the improved C4.5 algorithm,three data sets in UCI were selected and the results showed that the improved algorithm was effective and feasible.In this paper,the customer value and the business type of two-dimensional retail bank customer precision classification model are constructed with the improved C4.5 algorithm.According to this model,the corresponding classification rules are derived.After testing the classification model,its classification accuracy reached 82.66%,which basically meets the actual needs.To make this accurate classification model of retail banking customers applied to large-scale data sets,we parallelize the algorithm used in this model on the Hadoop platform.