Supply Chain Optimization of Industrial Machinery Projects Based on Artificial Neural Networks for Multi-Product Demand Forecasting

Fluctuating demand and diverse products pose significant challenges for supply chain management in manufacturing, leading to production disruptions and increased costs. This study aims to optimize the supply chain of the industrial machinery sector through demand forecasting with artificial neural networks (ANN). Demand data covering 24 months for three products—water tanks, mixer machines, and conveyors—were analyzed using time series clustering based on autocorrelation and complete linkage. Stationarity was tested using the Augmented Dickey-Fuller (ADF) and Box-Cox transformation prior to modeling. The ANN was trained with a learning rate of 0.001, 20 hidden neurons, and PACF-based lag inputs. Results showed that cluster one achieved a MAPE below 30%, indicating strong predictive accuracy, while cluster two achieved a MAPE below 50%, considered acceptable. Predictions for the 25th month from cluster one are recommended for inventory planning. The study demonstrates that ANN-based forecasting effectively supports decision-making and enhances supply chain optimization in multi-product manufacturing.