Customers are a critical asset to a company's success and ensuring their satisfaction is paramount. However, continuous churn can lead to reduced value flowing from customers, potentially jeopardizing a company's competitive advantage. Customer churn, where consumers choose products from other brands, is influenced by various factors such as promotion, price, product availability, and customer satisfaction levels. While much of the research on churn prediction is concentrated in the telecommunications, retail, and banking industries and only a few have conducted churn prediction research on online stores. This research aims to utilize data mining with a focus on machine learning algorithms, especially the tree-based gradient boosted models method that applies XGBoost, LightGBM, and CatBoost models, to predict customer churn in online stores. The research methodology involves data collection, data pre-processing, model selection and training, model evaluation, analysis and results. This research uses several libraries such as pandas library, numpy, matplotlib, and so on. The results of this study show that the XGBoost model achieved the highest accuracy in predicting customer churn, with an ROC curve of 0.66 and an accuracy value of 0.80032. The feature importance analysis highlights the gender variable as an important factor in model performance. This research contributes to improving customer service, minimizing churn, and ultimately increasing company profitability in the online store sector. Suggestions for future research include expanding data sources, testing with more evaluation metrics, exploring additional churn factors and comparing with other prediction methods for validation.
Copyrights © 2024