Evaluating the fidelity of statistical forecasting and predictive intelligence by utilizing a stochastic dataset

Forecasting plays a pivotal role in reducing waste within the manufacturing enterprise by aligning production closely with demand, thereby minimizing the excess that typically results from overproduction. Accurate forecasting models enable decision makers to predict demand with greater precision, which leads to several waste-reducing benefits such as inventory optimization, improved resource allocation, enhanced supply chain coordination, and overall quality improvements. Industry 4.0 has catalyzed the convergence of traditional production and manufacturing methodologies with advanced predictive analytics, creating a paradigm shift in production systems. By utilizing a stochastic dataset and various performance measurements, this paper aims to evaluate the fidelity of traditional statistical forecasting algorithms and the newfound algorithms in Machine Learning and Deep Learning models. Our findings suggest that the Recurrent Neural Networks model outperformed all models when n lags = 10, while the Neural basis expansion analysis for the interpretable time series forecasting model outperformed all models when n lags = 30. Thus, it enhances the efficiency and agility of manufacturing operations and provides data-driven insights that can facilitate continuous improvement.