Optimizing Customer Retention with LightGBM: A Machine Learning Approach

Customer retention is one of the cornerstones for bringing about sustained growth in e-commerce.This study aims toward enhancement of customer retention prediction through LightGBM-a highly efficient supervised machine learning algorithm-using the Sales Data FY 2020 to 2021 dataset.The dataset hence comprises 286,392 records with 36 attributes, including customer demographics and transaction histories.LightGBM was chosen for its ability to handle largescale, high-dimensional data efficiently while addressing class imbalance.The methodology involved data preprocessing, feature engineering, and model training.Hyperparameter tuning was applied, and LightGBM's performance was evaluated using accuracy and F1-score.Results show that LightGBM outperforms Random Forest and AdaBoost in both accuracy and processing speed, with monetary value and transaction frequency being key factors influencing retention.This study provides a scalable CRM system enabling businesses to enhance retention strategies through targeted marketing and personalized engagement.