<![CDATA[This study aims to develop a predictive model for forecasting product sales using the Multilayer Perceptron (MLP) algorithm. The model's performance was evaluated using key metrics, including the Mean Absolute Error (MAE), Mean Squared Error (MSE), and R² score. The model achieved an MAE of 0.861, an MSE of 9.521, and an impressive R² score of 0.999, demonstrating its ability to accurately predict product sales with minimal error. Feature correlation analysis identified key variables related to the target prediction, which is the number of products ready for shipment, underscoring the importance of feature selection in enhancing model performance. Prediction results revealed variability among product sales, with products like Foodpak Matte 245 (Code 49) predicted to sell approximately 244.31 units, while others like Stiker Kertas (Code 90) showed lower sales forecasts. The findings suggest that strategic interventions may be necessary to boost sales for underperforming items and capitalize on the demand for popular products. Future improvements, such as optimizing the network architecture, experimenting with activation functions and optimization algorithms, and incorporating external factors such as market trends, could further enhance the model’s accuracy and predictive power. Overall, the MLP model demonstrates strong potential for product sales forecasting, providing valuable insights for business decision-making.]]>
Copyrights © 2025
