A Machine Learning Evaluation Method for Sustainability Evaluation: The Case of Neighbourhoods' Design

This paper proposes a framework for machine learning to evaluate landscape design. In this study, we measured key performance indicators of landscape-development plans using a convolutional neural network (CNN) approach to predict the performance level of the design. The model used 3749 performance evaluations from 36 professionals, covering six sustainability criteria in 32 neighbourhoods' designs. Results show a high agreement level between experts on the performance level of the designs. The study contributes to computational sustainability by showing the potential in evaluation-automation of urban resiliency, ecological enhancement, and design for wellbeing, using expert knowledge and machine learning.