PENERAPAN METODE DATA MINING UNTUK PREDIKSI CUSTOMER CHURN PADA PT XYZ

• Main Author: Jefri, Bobby
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/78731
• Institution: Institut Teknologi Bandung

PT XYZ is a company operating in the retail industry and has adopted an e-commerce business model to remain competitive. The growth of internet users in Indonesia is an indicator that there is potential market development for the e-commerce business model, which requires users to have internet access. PT XYZ has an average churn rate of 20.25%, which is higher than the industry's average churn rate of 7.55%. Predicting customers with a high probability of churning can assist in implementing effective retention strategies. This research is conducted by applying the Cross Industry Standard Process for Data Mining (CRISP-DM) data mining methodology to predict customer churn. The alternative models used for this research are classification-based decision tree models, including Random Forest, XGBoost, and CatBoost. The best model is selected based on the performance evaluation of prediction results. From the conducted research, the best predictive model is Random Forest, with hyperparameters: n_estimator = 150, max_depth = 20, min_sample_split = 2, and max_features = sqrt(3.32). This model has a sensitivity/recall of 99.22%. The best predictive model serves as the core in designing a prototype application for predicting customer churn. The churn prediction application prototype is created using the Python programming language in the form of a Graphical User Interface (GUI).