Extracting Features from Optimizsation Techniques to Discover Liability of Customers Using Decision Trees

Abstract

Most algorithms based on data mining are used to discover customer models for distributing information, which is used in Customer Relationship Management (CRM), for pointing out customers who are loyal and who are attritors, but human expertise is a must for discovering knowledge manually. Many post processing techniques have been introduced that do not suggest action to increase the objective function such as profit. In this paper, a feature extraction technique is proposed for the best approximation property. An Uncorrelated Discriminant Analysis (UDA) algorithm based on Maximum Margin Criterion (MMC) is used. The extracted features via UDA are statistically uncorrelated. It serves as an effective solution for small sample size problem. The extracted features are given as an input to a novel algorithm that suggests actions to change the customer from the undesired status to the desired one. These algorithms can discover the reduction in cost and transform customer from undesirable classes to desirable ones. The UDA algorithm is evaluated in terms of classification accuracy and robustness. Many tests have been conducted and experimental results have been analyzed in this paper.