Prediction of human restorative experience for human-centered residential architecture design: A non-immersive VR–DOE-based machine learning method

Abstract

Nowadays, the topic of restorative experience in built environments has attracted more attention because of the increasing stress levels in modern society. Researchers have sought to identify the architectural features that influence a person's perceived restorative experience to achieve human-centered architectural designs. However, the relevant design knowledge is unsystematically scattered, making it difficult for designers to interpret information and make informed decisions in practice. This paper explores the feasibility of machine learning in capturing the restorative quality of design alternatives, thereby providing decision support for proactive architectural design analysis. To deal with feature selection and the uncertainty associated with affective modeling, a framework is introduced that integrates design of experiments and machine learning methods. The human restorative experience is assessed within non-immersive VR environments using self-reported psychometric scales. Consequently, general regression neural network is revealed as superior to other machine learning methods in forecasting the restorative experience.