ARIMA versus artificial neural network for Thailand’s cassava starch export forecasting

• Main Authors: Warut Pannakkong, Van Nam Huynh, Songsak Sriboonchitta
• Format: Book Series
• Published: 2018
• Subjects: Computer Science
• Online Access: https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84952684543&origin=inward; http://cmuir.cmu.ac.th/jspui/handle/6653943832/55588
• Institution: Chiang Mai University

© Springer International Publishing Switzerland 2016. Thailand is the first rank cassava exporter in the world. The cassava export quantity from Thailand influences cassava trading in international market. Therefore, Thailand’s cassava export forecasting is important for stakeholders who make decision based on the future cassava export. There are two main types of cassava export which are cassava starch and cassava chip. This paper focuses on the cassava starch, which is around 60 % of the total cassava export value, including three following products: native starch, modified starch and sago. The cassava starch export time series from January 2001 to December 2013 are used to predict the cassava starch export in 2014. The objectives of this paper are to develop ARIMA models and the artificial neural network (ANN) models for forecasting cassava starch export from Thailand, and to compare accuracy of the ANN models to the ARIMA models as benchmarking models. MSE, MAE and MAPE are used as accuracy measures. After various scenarios of experiments are conducted, the results show that ANN models overcome the ARIMA models for all three cassava starch exports. Hence, the ANN models have capability to forecast the cassava starch exports with high accuracy which is better than well-known statistical forecasting method such as the ARIMA models. Moreover, our finding would give motivation for further study in developing forecasting models with other types of ANN models and hybrid models for the cassava export.