PEMBANGUNAN MODEL CUSTOMER CHURN PADA PT XYZ MENGGUNAKAN METODE DATA MINING

• Main Author: Joseph Billie C, Johannes
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/71075
• Institution: Institut Teknologi Bandung

PT XYZ is a company that was established in 2020 and specializes in creating financial management applications. However, the new application was publicly launched in late 2021. As of April 2022, PT XYZ has a total user count of around 22,000 people. From January to March 2022, it was found that the average churn rate of the application was 63%. The high churn rate was obtained partly due to the ineffective customer retention method. The company has not been able to process historical data to see the users' application behavior patterns. Therefore, this research is conducted to make machine learning models and prototype that can be used by the company to predict the churn behavior of the users. This research refers to the Cross Industry Standard Process for Data Mining (CRISP-DM) methodology standards. The process stages carried out with this method are business understanding, data understanding, data preparation, modeling, evaluation, and deployment. The alternative algorithms used to make machine learning are Adaboost, XGBoost, Catboost, and Random Forest. The model and prototype development are done using the Python programming language. From the research that has been done, the Random Forest model is the best model for predicting customer churn for PT XYZ's application users with an accuracy of 81.3% and F1-Score of 81.3%. The five most significant variables for the model are balance, monthly_income, avg_weekly_spending, avg_weekend_spending, and sum_Lainnya_o. Then the model is implemented in a web-based prototype application using the Streamlit library.