CUSTOMER CHURN PREDICTION MODEL DEVELOPMENT FOR COMPANY X USING DATA MINING TECHNIQUE

• Main Author: H A, Iffatabiyan
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/79368
• Institution: Institut Teknologi Bandung

PT X is a private banking company in Indonesia that has been operating since 1959. From July 2020 to January 2021, PT X has 12,73% churn rate. This relatively big percentage was obtained because of the poor detection method that PT X currently had, which does not utilize historical data to see customer behavioral patterns before they leave the company. Currently, PT X does not have an information system to learn customer behavioral patterns before they churn. This research is conducted to build a potential customer churn prediction model using historical data owned by PT X. The Cross-Industry Standard Process Method for Data Mining (CRISP-DM) is used as a framework to build the predictive model. This research used customer data obtained from the Data Analytics Division with thirty-five variables as a start. Two alternative algorithm, Random Forest and Gradient Boosting, were used to build the model with Python as its programming language. From the conducted research, the best model obtained was Random Forest with hyperparameter settings on max_depth, max_features, min_sample_split, and n_estimators respectively are 15, auto, 5, and 100. EOP_IDR, OTH_SAV, TAB_REG, VINTAGE, and SPKIDS are the top five variables with high importance for the model. The model has 72.81% of accuracy and 86,7% recall rate. The model then implemented in the form of Graphical User Interface using Python as its programming language.