Improving the Performance of Forecasting Models with Classical Statistical and Intelligent Models in Industrial Productions

The capability to receive and deliver customer demand on time in today's competitive world is a significant concern for all industries. In particular, demand management has entered a new era with many companies competing in the last decade. Customer demand management is one of the contemporary issues. The main goal of demand management is to improve supply chain effectiveness, and it is important to note that it is complementary to distribution management and product demand management. Therefore, demand forecasting is essential. For this purpose, in this study, the modeling of the combination structure using autoregressive integrated moving average models and multilayer perceptron neural networks in the field of demand with benchmark data is investigated. The data sets used in this study are two well-known benchmarks of the total product revenue of the Taiwan machinery industry and the sales volume of soft drinks. Eviews and Matlab software have been used to determine the unknown parameters of the proposed model. The experimental results of the research show that the performance of the proposed hybrid model is more accurate than its single components. In addition, results indicate that intelligent models can perform better than classic statistical models.

Keywords:

Seasonal Autoregressive Integrated Moving Average (SARIMA) , Seasonal Artificial Neural Network (SANN) , Demand Forecast