Supervised Machine Learning Models to Assess Impact of Building Parameters on Energy Efficiency

It is important to maintain higher energy efficiency of a building to save energy. One way to achieve this is by reducing heating load (HL) and cooling load (CL) and they are significantly impacted by parametric building design. Reduced HL and CL define a higher energy efficiency of a building. Towards this, we studied the impact of design input parameters: relative compactness, surface area, wall area, roof area, overall height, orientation, glazing area, and glazing area distribution, on the output variables: heating load and cooling load with a 768-building data obtained from UCI machine learning repository. Their impact was analyzed using model coefficients obtained by Logistic regression and Linear regression with 10-fold cross validation. Visualization using Word-Cloud supported easy understanding of observations. Linear regression predictions obtained results with root mean square error: 2.82, 2.13 and mean absolute error: 1.97, 2.13 and Logistic regression was able to achieve 76.30% and 73.17% accuracy for HL and CL. The analysis of coefficients indicated that reduced overall height, reduced glazing area, and increased relative compactness value can achieve a higher energy efficiency of a building and it can be applied to the real-world to build energy efficient structures.