Role of Artificial Intelligence (AI)-Driven Demand Forecasting: A Machine Learning Approach for Supply Chain Resilience

Reducing financial risks, improving inventory management, and strengthening supply chain resilience all depend on accurate demand forecasts. Traditional forecasting methods often struggle with unpredictable market fluctuations, seasonal variations, and external disruptions, leading to inefficiencies such as stockouts and overstocking. This study leverages artificial intelligence (AI) and machine learning techniques to improve sales prediction accuracy using real-world Walmart sales data. This study utilizes ML techniques to predict sales accurately, comparing XGBoost, LightGBM, Random Forest, and K-Nearest Neighbors (KNN). A methodology involves data preprocessing, including data cleaning, one-hot encoding, and normalization, followed by feature selection and dataset splitting. XGBoost and LightGBM models outperform traditional methods, achieving high R2 values of 0.9752 and 0.9732, respectively, with low MSE, RMSE, and MAE, indicating strong predictive capabilities. Comparative analysis reveals that Random Forest (R2 = 0.9569) and KNN (R2 = 0.9381) exhibit lower accuracy. The actual vs. predicted sales plots for XGBoost and LightGBM demonstrate close alignment, while residual plots confirm minimal bias. Overall, the findings highlight the superiority of gradient boosting techniques in demand forecasting, offering valuable insights for effective sales prediction and inventory planning in the retail sector.