Predicting energy load of L-shaped courtyard buildings in Malaysia using machine learning

Mohd Sabri, Alea Syaffa and Hassan, Raini (2026) Predicting energy load of L-shaped courtyard buildings in Malaysia using machine learning. International Journal on Perceptive and Cognitive Computing (IJPCC), 12 (2). pp. 118-128. E-ISSN 2462-229X

PDF - Published Version Restricted to Repository staff only Download (511kB) | Request a copy

Abstract

Building’s account for a substantial share of global energy consumption, with cooling demand dominating in hot and humid tropical climates such as Malaysia. Although dynamic simulation tools such as EnergyPlus and TRNSYS provide accurate performance evaluation, their high computational cost and reliance on specialised expertise limit their application during early-stage design. This study develops a machine learning-based surrogate modelling framework to predict the cooling load, heating load, and total energy load of L-shaped courtyard buildings in Kuala Lumpur. A validated parametric simulation dataset was generated using key architectural and geometric variables, including WIDTH, LENGTH, HEIGHT, ORIENTATION, WINDOWS_RATIO, FORM_FACTOR, and S_V_RATIO. Three supervised regression models, Random Forest (RF), Gradient Boosting (GB), and Artificial Neural Network (ANN-MLP), were evaluated using Root Mean Square Error (RMSE) and Coefficient of Determination (R²). The results show that ensemble learning models consistently outperform ANN, with Gradient Boosting achieving the highest predictive accuracy across all energy outputs. For cooling load prediction, GB achieved an RMSE of 0.0188 and an R² of 0.9997, while also producing the best performance for heating load and total energy load. The findings further indicate that envelope-related variables, particularly the window-to-wall ratio, have a significant influence on cooling and overall energy demand in tropical courtyard buildings. By integrating climate-specific parametric simulation with ensemble machine learning, the proposed framework accurately reproduces simulation-derived energy loads while reducing the need for repeated simulation during early-stage design. The developed surrogate model provides a reliable decision-support tool for rapid performance assessment during early-stage architectural design and contributes a validated predictive framework for energy-efficient tropical courtyard buildings in Malaysia.

Item Type:	Article (Journal)
Uncontrolled Keywords:	Building energy prediction, courtyard buildings, tropical climate, machine learning, Gradient Boosting
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Kulliyyahs/Centres/Divisions/Institutes (Can select more than one option. Press CONTROL button):	Kulliyyah of Information and Communication Technology > Department of Computer Science Kulliyyah of Information and Communication Technology > Department of Computer Science
Depositing User:	Dr. Raini Hassan
Date Deposited:	09 Aug 2026 11:36
Last Update:	09 Aug 2026 11:36
Queue Number:	2026-08-Q4536
URI:	http://irep.iium.edu.my/id/eprint/130551

Actions (login required)

View Item