Comparison of Model Performance between Penalized Regression and Machine Learning for Diesel Prices in Thailand

Abstract

This research study was to compare of model performance for the price of diesel B7, diesel B10, and diesel B20 and find a suitable model to forecast the price of diesel B7, diesel B10, and diesel B20 using daily data from January 1, 2020 to December 31, 2022. The factors used in the study were crude oil prices in the world market, consumption of diesel B7, diesel B10, and diesel B20, exchange rate, consumer price index, oil fuel fund rate of diesel B7, diesel B10, and diesel B20, and crude palm oil prices. Then, compare models to find the best model by measuring the model’s performance with the Root Mean Square Error (RMSE) using Stepwise Regression, Penalized Regression such as Ridge Regression, Lasso Regression, Adaptive Lasso Regression, Elastic Net Regression, and Machine Learning such as Support Vector Regression and Random Forest. According to the research results, Random Forest is the most suitable method for forecasting the price of diesel B7, diesel B10, and diesel B20 with RMSE values of 0.379, 0.3833, and 0.3539, respectively.