Reinforced and Unreinforced Machine Learning Application in Sustainable Building Design and Performance
Document Type : Original Research
Authors
1
Department of Architecture, Ma. C., Islamic Azad University, Mashhad, Iran
2
Assistant Professor, Department of Civil Engineering, Faculty of Engineering, Golestan University, Gorgan, Iran
10.48311/jaert.2026.123626.82814
Abstract
ABSTRACT
Architecture has traditionally relied on qualitative research, but data-driven design is gaining prominence as data becomes more accessible. Machine Learning Algorithms (MLAs) have been effectively applied to building design, demonstrating their potential to enhance building performance across numerous sectors. This article provides a comprehensive overview of Machine Learning (ML) applications in building design, covering architectural design and visualization, material optimization, construction management, smart operations, life cycle analysis, and the circular economy. It investigates how ML can be applied at every stage of the building lifetime, from design to upkeep.
The results show that ML can support environmental sustainability, boost occupant well-being, and improve building performance. The paper introduces fundamental ML concepts, data-driven processes, and the categorization of algorithms—supervised, unsupervised, and reinforcement learning—along with widely used methods such as Artificial Neural Networks, Deep Neural Networks, Convolutional Neural Networks, Support Vector Machines, Generative Adversarial Networks, Variational Autoencoders, Decision Trees, and Random Forests. It also discusses key challenges, including the need for balanced and large datasets, low model explainability, and difficulties in generalizing predictions when data distributions shift during early design stages.
The findings indicate that ML can facilitate environmental sustainability, enhance occupant well-being, and optimize building performance. A path for architects and construction professionals to successfully integrate ML and data science into their practices is also provided, along with identified future trends and research opportunities.
Architecture has traditionally relied on qualitative research, but data-driven design is gaining prominence as data becomes more accessible. Machine Learning Algorithms (MLAs) have been effectively applied to building design, demonstrating their potential to enhance building performance across numerous sectors. This article provides a comprehensive overview of Machine Learning (ML) applications in building design, covering architectural design and visualization, material optimization, construction management, smart operations, life cycle analysis, and the circular economy. It investigates how ML can be applied at every stage of the building lifetime, from design to upkeep.
The results show that ML can support environmental sustainability, boost occupant well-being, and improve building performance. The paper introduces fundamental ML concepts, data-driven processes, and the categorization of algorithms—supervised, unsupervised, and reinforcement learning—along with widely used methods such as Artificial Neural Networks, Deep Neural Networks, Convolutional Neural Networks, Support Vector Machines, Generative Adversarial Networks, Variational Autoencoders, Decision Trees, and Random Forests. It also discusses key challenges, including the need for balanced and large datasets, low model explainability, and difficulties in generalizing predictions when data distributions shift during early design stages.
The findings indicate that ML can facilitate environmental sustainability, enhance occupant well-being, and optimize building performance. A path for architects and construction professionals to successfully integrate ML and data science into their practices is also provided, along with identified future trends and research opportunities.
Keywords
Subjects
