Predicting energy performances of buildings' envelope wall materials via the random forest algorithm

Purpose: Numerous simulation software has been used to evaluate energy performance with 12% of the research focusing on long-term energy consumption prediction. This paper aims to utilize machine learning to predict the energy performance of building envelope wall materials over extended periods. Methodology: In our work, machine learning model learns from a large set of building envelopes simulated using the Integrated Environmental Solutions Virtual Environment as follows: Findings: Machine Learning models can also be used to predict the impact of building design and construction characteristics on energy consumption, showing that factors such as wall thickness, orientation, and thermal mass indicated lower relative standard error (<0.001); however, not all of them were statistically significant (p > 0.05). While the overall model indicates statistical significance (p = 2e-16), the multivariate linear regression model produces R2 value of 0.42, indicating a weak relationship between predictor variables and target attributes. Originality: The utilisation of Random forest algorithm for the wall envelop energy consumption Research implecation: different to other techniques, our proposed approach addressed the issue related to building envelop for new constructions to assist professional from construction industry.