A COMPARATIVE ANALYSIS OF FINE TREE REGRESSION AND ANFIS FOR PREDICTING CARBON FOOTPRINTS IN RESIDENTIAL CONSTRUCTION

Abstract

The construction industry is a major contributor to global carbon emissions, necessitating accurate predictive models for sustainable development. This study compares the performance of fine tree regression (Rtree) and adaptive neuro-fuzzy inference system (ANFIS) in predicting carbon footprints across four stages of residential construction: production, transportation, operational, and destruction. A dataset of 2000 observations was used, with an 80 to 20 split for training and testing. The models were evaluated using root mean square error (RMSE) and mean absolute percentage error (MAPE). Results indicate that ANFIS outperforms Rtree in all stages. ANFIS achieved an RMSE of 0.5142 at the production stage compared to 0.5317 for Rtree. ANFIS obtained an RMSE of 447.07 in the transportation stage, while Rtree recorded 492.23. The operational stage showed an RMSE of 1179.3 for ANFIS versus 1386.5 for Rtree. At the destruction stage, ANFIS demonstrated superior accuracy with an RMSE of 0.0610 compared to 0.0631. The findings suggest that ANFIS provides more precise predictions and is a reliable model for estimating carbon footprints in residential construction. This study contributes to sustainable construction by offering an efficient tool for reducing environmental impact.