Applied Machine Learning Algorithms for Courtyards Thermal Patterns Accurate Prediction

Currently, there is a lack of accurate simulation tools for the thermal performance modeling courtyards due to their intricate thermodynamics. Machine Learning (ML) models have previously been used predict and evaluate structural buildings as means solving complex mathematical problems. Nevertheless, microclimatic conditions building surroundings not thoroughly addressed by these methodologies. To this end, in paper, adaptation ML techniques more comprehensive methodology fill research gap, covering only prediction courtyard microclimate but also interpretation experimental data pattern recognition, proposed. Accordingly, based on climate zoning aspect ratios 32 monitored case studies located South Spain, Support Vector Regression (SVR) method was applied measured temperature inside courtyard. The results provided strategy showed good accuracy when compared data. In particular, two representative studies, if daytime slot with highest urban overheating considered, relative error almost below 0.05%. Additionally, values statistical parameters are agreement other literature, which use computationally expensive CFD show than existing commercial tools.