Thoranin Sujjaviriyasup

DOI: 10.14456/sjst-psu.2018.22

pp. 75 - 78

Abstract

In this research, a combined approach emphasizing on data pre-processing technique is developed to forecast prices of
agricultural commodities in Thailand. The future prices play significant role in decision making to cultivate crops in next year.
The proposed model takes ability of MODWT to decompose original time series data into more stable and explicit subseries, and
SVR model to formulate complex function of forecasting. The experimental results indicated that the proposed model
outperforms traditional forecasting models based on MAE and MAPE criteria. Furthermore, the proposed model reveals that it is
able to be a useful forecasting tool for prices of agricultural commodities in Thailand.

Keywords

• combined approach
• agricultural commodity price
• discrete wavelet transform
• ARIMA
• support vector regression