Based Tree Algorithm On Bank Customer Segmentation Research

Along with the development of banking,bank customer information increase sharply,for better for Banks to provide effective financial services,the bank practitioners need to subdividing customer and divide customers into different categories according to the customer characteristics,and then formulated for different categories of customer service strategy.Tree algorithm in data mining field in recent years,much attention has been paid to gradually become a hot academic research field.In addressing the problem of classification tree algorithm has many advantages,including can effectively handle a large number of high dimensional and complex data structure,the outliers and noise point has strong tolerance,can be more intuitive interpretation of results,etc.,is the most commonly used CRM customer segmentation technology.This paper first discusses the bank customer segmentation problems and the research significance,the comprehensive analysis of the importance of customer segmentation,this article provides the theoretical support;Then discussed based on the distributed computing tools,Spark memory,Spark in dealing with a large amount of data has very obvious advantages;Finally,focus on the four tree algorithm in the application of bank customer segmentation problem,details the CART decision tree C4.5 algorithm and decision tree algorithm,the principle of balance with random forests BRF GBDT algorithm,and the GBDT algorithm in the final result is improved on the question of weighting,respectively dealing with the bank customer data and results,according to each tree algorithm's performance in the experiment were analyzed,and the final balance study random forests performed best in the bank customer segmentation.In this paper,the random forest algorithm is introduced into the bank customer segmentation,and its ability to predict by empirical testing;at the same time explore the distributed computing platform Spark,solve the problem of large amount of data;finally the GBDT algorithm is improved on weight problem.