Incorporating Qualitative Metrics into a Computational Model using Machine Learning

Laura Kaersaa, Poul Henning Kirkegaard

Abstract: Within a computational model, quantitative metrics frequently steer the decision-making process, where qualitative metrics, including elements representing the architectural expression, are often neglected. The motivation behind this article is to quantify the elements that establish the harmony between form, material and technique, resulting in a combined tectonic design. This article aims to illustrate how a computational model can include qualitative architectural information and thereby create an honest tectonic design. The most influential theories concerning architectural values have been examined. Based on this, six qualitative values are defined as: symmetry, elements, texture, material placement, scale and variety which describe the architectural value, or aesthetic quality, of a space. Based on the defined qualitative values, a machine learning model has been defined, which is based on a pseudorandom sampling plan and a shallow artificial neural network. A computational model has been developed, implemented with the machine learning model predicting aesthetic quality. An optimisation process is defined, utilising a genetic algorithm to conduct a meta-heuristic optimisation. From this computational model, a case study has been conducted which considers a structural objective, an acoustic objective and an aesthetic constraint. The solution space of the case study has been examined in a generative design process. The results displayed the possibilities of informing a computational model with aesthetic quality while maintaining acoustic and structural integrity. The research conducted in this article presents a novel framework for creating tectonic designs through a computational model.

Keywords: Tectonic design, Aesthetic Metrics, Computational architecture, Machine Learning, Architectural modeling

How to Cite?: Laura Kaersaa, Poul Henning Kirkegaard, "Incorporating Qualitative Metrics into a Computational Model using Machine Learning", Volume 14 Issue 11, November 2025, International Journal of Science and Research (IJSR), Pages: 236-247, https://www.ijsr.net/getabstract.php?paperid=SR251021004841, DOI: https://dx.doi.org/10.21275/SR251021004841