BANK CUSTOMER CHURN PREDICTION MODEL: THE CASE OF COMMERCIAL BANK OF ETHIOPIA

Abstract

In 21st century because of availability and affordability of computer technology, organizations and businesses especially in banking sector are situated in basic requirement to gain a number of key advantages to improve their business using Machine Learning (ML) Algorithm. ML Algorithms is a branch of artificial intelligence based on the idea that systems can learn from data, identify model and make to support decision with minimal human intervention brief about the customer churn. Nowadays industries working with large amounts of data have recognized the value of machine learning in this case Commercial bank of Ethiopia (CBE). CBE is one of such service-giving industries that collects, processes and stores huge amounts of records from time to time and therefore deal with large amount of data. On the other hand, CBE is facing problems in Customer Relationship Management (CRM), specifically it is unable to control the customer churn. Customer Churn is the propensity of a customer to stop doing business with an organization and subsequently moving to some other company. In this study an attempt is made to apply machine learning algorithms for customer churn prediction. After performing business and data understanding the data preparation task is done to clean and make the data ready for experimentation. For the experiment and construct predictive model, machine learning algorithms such as SVM, KNN, Naïve Bayes and Logistic Regression are selected based on their advantages and past performance seen in different literatures, it has been reported that they were widely used classifier algorithms for prediction and classification. The R Studio with R programming was used to simulate all the experiments. Confusion matrix was used to calculate the accuracy, recall and precision and evaluate the performance of the models. The results of the experiment show high accuracy, so that the models can be used to predict customer status accurately. Based on the research findings, the KNN classifier produced an accuracy of 99.91%, the SVM classifier produced an accuracy of 92.4%, Logistic Regression model also produced an accuracy of 93.8%, and Naïve Bayes classifier produced an accuracy of 83.8 %. Therefore, the KNN classifier is proposed for constructing bank customer churn prediction model for Commercial Bank of Ethiopia. Based on the proposed optimal model in this study, we recommend future research to integrate customer churn predictive model with CRM data base management system.