Customer churn forecast is a vital undertaking in the retail sector, necessary for maintaining company growth and improving customer loyalty. Customer churn prediction is crucial for company performance in retail settings because of the higher expense associated with acquiring new customers compared to maintaining existing ones. This research examines the use of machine learning techniques, especially a combination of Random Forest and XGBoost classifiers, to forecast customer attrition in retail settings. This research utilizes a comprehensive dataset including various client characteristics collected from a shopping mall. It takes a methodical methodology that includes data preparation, feature engineering, model training, hyperparameter optimization, and performance assessment. The use of Explainable Artificial Intelligence methods, such as SHapley Additive exPlanations (SHAP) and LSTM, guarantees the interpretability of the model without compromising its prediction accuracy. The solid predictive performance attained by the ensemble model created using soft voting provided useful insights into consumer behavior and churn dynamics. Furthermore, the clarification of SHAP values enables a detailed analysis of individual forecasts, providing insight into the fundamental factors that contribute to customer turnover. Additionally, the use of bee colony optimization yields more efficient outcomes. These results have practical significance for retail practitioners as they provide them with the opportunity to create focused strategies to retain customers and ensure the long-term viability of their firm. This research adds to the expanding collection of literature on customer churn prediction in the retail industry. It provides valuable information on the retail customer journey and helps in making well-informed decisions in a highly competitive marketplace.