Optimising L-shaped courtyard building design using prescriptive analytics

Mohd Sabri, Alea Syaffa and Hassan, Raini (2026) Optimising L-shaped courtyard building design using prescriptive analytics. International Journal on Perceptive and Cognitive Computing (IJPCC), 12 (2). pp. 111-117. E-ISSN 2462-229X

PDF - Published Version Restricted to Repository staff only Download (209kB) | Request a copy

Abstract

Buildings in tropical climates are predominantly cooling-dominated, making architectural design decisions an important factor in reducing operational energy demand. While machine learning has been widely applied to predict building energy performance, most existing studies focus primarily on predictive accuracy and provide limited guidance on translating predictions into practical design decisions. Consequently, architects and designers often know the expected energy performance of a design but remain uncertain about which configuration should be selected to achieve specific energy objectives. This study proposes a prescriptive analytics framework that extends machine learning beyond prediction to support architectural decision-making for L-shaped courtyard buildings in Malaysia's tropical climate. A Gradient Boosting (GB) model was identified as the best-performing predictive model, achieving R² values of 0.999684, 0.998808, and 0.999695 for cooling load, heating load, and total load prediction, respectively. The trained model was subsequently integrated into a prescriptive optimisation engine to identify energy-efficient architectural configurations. The optimal cooling- and total-load designs converged towards compact building forms with a window-to-wall ratio (WWR) of 0.10, producing predicted cooling and total energy loads of 43.81 kWh/m² and 45.55 kWh/m², respectively. In contrast, the heating-load optimum was achieved with a WWR of 0.40, resulting in a predicted heating load of 0.68 kWh/m². A scenario-based "what-if" analysis further confirmed that the cooling-oriented design package consistently achieved the lowest overall energy demand. The results identify the window-to-wall ratio as the most influential architectural parameter affecting building energy performance and demonstrate that cooling and heating objectives cannot be optimised simultaneously. The proposed framework transforms machine learning from a predictive model into a practical decision-support tool that provides interpretable and actionable design guidance for architects during the early stages of building design.

Item Type:	Article (Journal)
Uncontrolled Keywords:	Prescriptive Analytics, Machine Learning, Building Energy Prediction, Gradient Boosting, Courtyard Buildings
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 09:57
Last Update:	09 Aug 2026 11:37
Queue Number:	2026-08-Q4534
URI:	http://irep.iium.edu.my/id/eprint/130548

Actions (login required)

View Item