Enhanced E-Commerce Demand Prediction through Ensemble Models and Optuna-Based Hyperparameter Optimization

The fast growth of e-commerce has made sophisticated prediction models necessary to precisely forecast demand and maximize inventory control developed. In a very competitive industry, effective demand forecasting is absolutely essential for lowering costs, avoiding waste, and raising customer satisfaction. This work presents a complex ensemble-based method combining Random Forest models with hyperparameter optimization enabled by Optuna with LightGBM and XGBoost. This approach guarantees best performance via exact tweaking and makes use of the features of every model. Rigorously evaluated on actual e-commerce data, our suggested model shows a notable lower MAPE than independent models. With a MAPE of 0.0341, the ensemble model specifically outperformed LightGBM (0.0344), XGBoost (0.00336), and Random Forest (0.0358) by roughly 2–5%. This development emphasizes how well combined ensemble learning with hyperparameter optimization increases predicted accuracy. Furthermore, our method provides scalability and adaptability, which qualifies for many e-commerce situations from high-dimensional data settings to perishable items. This study shows that the suggested ensemble model offers a strong, adaptable demand forecasting solution for e-commerce and increases forecast accuracy, therefore opening the path for more effective and data-driven decision-making procedures.