An application of the Random Forest algorithm for the prediction of Solar Envelope ‘Floor Space Index’ based on spatiotemporal parameters

Abstract

It is becoming increasingly important to consider access to the sun in the early stages of a new building's design process. A Solar Envelope is an architectural form that meets required levels of density, whilst also assuring solar access to neighbourhood surfaces, thereby making it feasible to implement active and passive solar technologies there. This preliminary study investigated the cross-cutting intersection between urban morphology, Solar Envelope forms, and a Random Forest algorithm to predict the Solar Envelope's Floor Space Index. Recognizing that Solar Envelope forms depend on various spatiotemporal contexts, mainly the solar exposure cut-off period (equinoxes/solstices) and various neighbourhood geometrical attributes, simulations varying only the rooftop typologies (flat, gable, hip, sled, mansard, and pyramid) and façade orientations (south, east, west, and north) were conducted to evaluate how solar access to those surfaces affect the Floor Space Index of the Solar Envelope. Architectural tools were used to conduct simulations reflecting varying spatiotemporal scenarios. The data was extracted and represented geometrically, involving polar loci, and this synthetic dataset was subsequently utilized to train and evaluate a Random Forest algorithm. The results of this study show that the Random Forest algorithm was able to predict the Floor Space Index of the Solar Envelope with about 90.18% accuracy, though the only parameter influencing the regression model was the solar exposure cut-off period and not any other morphological features. Therefore, the study sheds insight into the need for the development of a more robust dataset through the simulation of randomized and diversified neighbourhood morphologies.