Muhammad Farid, Teuku Muhammad Faiz Nuzullah, Zatul Aklya, Syifa Nazila, Muhammad Zia Ulhaq , Feby Apriliansyah, Novi Reandy Sasmita

Optimizing Energy Consumption Prediction Across the IMT-GT Region Through PCA-Based Modeling

Introduction

Optimizing energy consumption prediction across the imt-gt region through pca-based modeling. Optimize IMT-GT energy consumption prediction using a PCA-Random Forest model. It reduces multicollinearity, boosting R2 to 0.993 and cutting errors for reliable, sustainable energy forecasting.

Abstract

This study aims to improve the accuracy of energy consumption prediction in the Indonesia-Malaysia-Thailand Growth Triangle (IMT-GT) region by addressing multicollinearity among independent variables such as energy production (Mtoe), lignite coal production (million tons), crude oil production (million tons), refined oil production (million tons), natural gas production (billion cubic meters), and electricity production (terawatt-hours). By integrating Principal Component Analysis (PCA) with Random Forest (RF), six correlated variables were reduced into two uncorrelated principal components (PC1 and PC2), explaining 80.77% of the data variance. The PCA-RF hybrid model outperformed the standalone Random Forest (RF) model, with an increase in the coefficient of determination (R2) from 0.976 to 0.993. Additionally, it achieved significant reductions in error metrics, with the mean absolute error (MAE) decreasing from 5.811 to 4.169 and the root mean square error (RMSE) dropping from 9.278 to 4.786. These results demonstrate PCA’s effectiveness in isolating dominant drivers such as energy and lignite coal production while improving model stability. The framework provides policymakers with a reliable tool to forecast energy demand and align economic growth with sustainability in fossil fuel-dependent economies.

Review

This study presents a compelling approach to enhancing energy consumption prediction within the critical Indonesia-Malaysia-Thailand Growth Triangle (IMT-GT) region. The core objective of mitigating multicollinearity among key energy production variables using a Principal Component Analysis (PCA) driven framework is highly relevant, given the inherent correlations in economic and energy datasets. The integration of PCA with a Random Forest (RF) model effectively addresses this challenge, reducing six correlated production variables into two principal components that collectively explain a substantial 80.77% of the data variance. This methodological choice is sound and promises more stable and interpretable models for a region vital to global energy dynamics. The performance gains achieved by the PCA-RF hybrid model are particularly impressive and represent a significant advance over the standalone Random Forest model. The increase in R2 from 0.976 to 0.993 indicates a near-perfect fit and robust predictive capability, while the substantial reductions in MAE (from 5.811 to 4.169) and RMSE (from 9.278 to 4.786) underscore the model's enhanced accuracy and reliability. The abstract highlights that PCA's effectiveness extends beyond mere dimensionality reduction, enabling the isolation of dominant drivers like energy and lignite coal production, which is a crucial insight for targeted policy interventions. These quantitative improvements make the proposed framework a powerful and trustworthy tool for energy forecasting. The practical implications of this research are significant, particularly for policymakers in fossil fuel-dependent economies within the IMT-GT region. A highly accurate and stable prediction model offers a strong foundation for strategic energy planning, enabling better alignment of economic growth with sustainability goals. While the abstract strongly demonstrates the model's superior performance, future work could further explore the implications of the remaining unexplained variance by the two principal components or consider incorporating external socio-economic or technological factors to broaden the model's scope. Nevertheless, this study makes a valuable contribution to the field of energy forecasting, providing a robust, data-driven approach that is both scientifically sound and practically applicable.

Full Text

Comments

You need to be logged in to post a comment.