Authors:   Pongsak Holimchayachotikul, Teresa Murino, Pachinee Payongyam, Apichat Sopadang, Matteo Savino, Romano Elpidio

Abstract

Owing to the U.S Hamburger crisis effect since the middle of 2007, frozen food products export industry sector, especially cooked chicken products export to Japan of Thai industry, endeavor has been spent in the supply chain management (SCM) of internal efficiency, merely aiming at competitiveness survival in terms of better quality and cost reduction. To reach the customer satisfaction, the company must work towards a right time and volume of his demand delivery. Therefore, forecasting technique is the crucial element of SCM operation. The more reducing inventory and capacity planning cost increase their company competitiveness; the more understanding how their company use the right forecasting based on information sharing in their SCM context. Currently, most of the companies, in this sector, do not have a right knowledge to implement the suitable forecasting system to sustain their business; furthermore, they only use top management judgment and some of the economical data for production. On the ground of the complex, stochastic, dynamic nature and multi-criteria of the logistics operations along the food products exporting to Japan of the Thai industry supply chain, the existing time series forecasting approaches cannot provide the information to operate demand from upstream to downstream effectively. The aim of the paper is to develop an innovative and simplified forecasting system and then implement for this industry based on data mining including time series factors and causal factors. This research methodology was designed with the case study company managers and engineers. Artificial neural network (ANN) theory was used to develop time series forecasting model for case study product. The type of ANN implemented was Multilayer Perceptron with the Quick-propagation training algorithm by using time series factor and causal factor such as Thai- Japan, EU-Thai and USA-Thai exchange rate, customer demand forecast and the other economic factors, from the case study company as input. The accuracy of the neural network model was compared with traditional customer demand forecast. The experimental results suggested that the ANN was capable of high accuracy modeling and resulted in much smaller error in comparison with the results from the present forecasting method of the company case study.