Forecasting Retail Oil and Natural Gas Vehicles Prices in Thailand Using Time Series Data Mining Techniques

Abstract

The purpose of this research is to develop the model of forecasting retail oil and natural gas vehicles prices for automobiles in Thailand using time series data mining techniques. There are three techniques such as Linear Regression, Multi-Layer Perceptron and Support Vector Machine for Regression. The data used for this study was collected the retail oil and natural gas vehicles prices in Thailand from 2012-2018 AD. totally 84 months. This research found that the suitable forecasting model for retail oil and natural gas vehicles prices as followed: 1) The forecasting model using Linear Regression was the most suitable for Gasohol E85 and Ultra Force Diesel, which had the rate of MMRE (Mean Magnitude of Relative Error) with the percentage of 2.46, and 4.60. 2) The forecasting model using Support Vector Machine for Regression was the most suitable for Gasohol 91, Gasohol 95, Gasohol E20 and Natural Gas Vehicles (NGV), which had the rate of MMRE with the percentage of 3.69, 3.20, 3.54, and 6.89, respectively.