Modelling the driving forces of green retrofitting adoption in developing nations: A data-driven approach

Abstract

<p>Green retrofitting (GR) has recently been witnessed as a practical strategy to lesser the negative<br>consequences of extreme energy use and carbon dioxide emissions of aged buildings. Accordingly,<br>many studies focused on the driving forces of GR adoption. Nonetheless, there seem to be limited<br>attempts on quantitative models to investigate the impacts of drivers for GR adoption, especially<br>in the developing economies context. Therefore, current research attempts to explore and model<br>the influence of drivers for GR adoption in the Sri Lankan developing economy. 12 semistructured<br>interviews and 122 questionnaire survey responses from the Sri Lankan context<br>were used to accumulate data. Partial least squares structural equation modeling (PLS-SEM) was<br>adopted to test the hypothesis concerning the influencing drivers on GR. Afterward, Machine<br>Learning algorithms were utilized to develop the prediction model. PLS-SEM results indicate that<br>building and occupant-related drivers,energy-related drivers, and environment-related drivers<br>have a noteworthy positive effect on Sri Lankan GR adoption. Moreover, the Random Forest<br>classifier was identified as the most optimal predicting model of the GR adoption perspective of<br>drivers with an accuracy score of 84 % and ROC AUC of 97.88 %. SHAP results advocate that<br>building and occupant-related drivers were significantly contributing to the model prediction.<br>Current research is significant in two ways. First, this is the very first attempt to predict the influences<br>of drivers for GR adoption in developing economies. Second, given the limited number of<br>quantitative analyses on GR adoption drivers, current research is worthwhile for policymakers<br>and industry practitioners in the decision-making process.</p>