Evaluation performance of time series methods in demand forecasting: Box-Jenkins vs artificial neural network (Case study: Automotive Parts industry)

Abstract

Production planning is a vital activity for manufacturing companies in managing any kind of organizational operations. One of the most important and vital tools that are considered necessary in production planning is forecasting future production demand. One of the most widely used approaches in demand forecasting is time series analysis. Also, two widely used methods in time series analysis are Box-Jenkins and Artificial Neural Network (ANN) approaches. In this study, the performance of these two methods to the types of errors and based on the concepts of multi-criteria decision making (MADM) has been investigated. These two methods are implemented for a product family in the automotive industry, and then the findings are compared and analyzed. The results showed that the Box-Jenkins method (Arima) provided much better predictions, which means that this method presented better results for 6 out of 8 products.

KEYWORDS:

• Arima
• Artificial Neural Network (ANN)
• Box-Jenkins
• demand forecasting
• seasonal Arima (Sarima)
• time series

Disclosure statement

No potential conflict of interest was reported by the author(s).