PREDICTING BANK CUSTOMER CHURN USING LOGISTIC REGRESSION AND DECISION TREES

• Main Author: Fachrizal Amni, Azka
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/81852
• Institution: Institut Teknologi Bandung

Banks, as vital financial institutions, play a crucial role in supporting economic growth and have experienced a shift in customer behavior towards digital services due to the Covid-19 pandemic. One strategy to address customer retention is by building a model that can predict bank customer churn using machine learning based on customer characteristics. This research develops logistic regression and decision tree models using customer data that includes demographic information, transaction patterns, and interactions with digital services. Logistic regression is a statistical method used to predict the probability of an event based on one or more independent variables, while decision trees are algorithms that split data into smaller subsets based on certain features to make predictive decisions. Among these models, the one with the best performance will be selected based on accuracy, precision, recall, F1-score, and the Receiver Operating Characteristic - Area Under Curve (ROC-AUC) metrics. The modeling results show that, within the data used, decision trees outperform logistic regression in predicting customer churn. The decision tree model produces a better evaluation from the confusion matrix and shows higher ROC-AUC values. Therefore, the decision tree model is chosen as the best model for bank customer churn analysis in this study.