Construction Waste Modeling for the Republic of Serbia

Abstract

The management of Construction Waste (CW) presents a significant challenge in sustainable development efforts. This study employs data modeling techniques to predict the annual quantities of various types of construction waste, encompassing total waste, metal waste, plastic waste, wood waste, mineral waste, and soil/concrete waste (CW1–CW8, respectively). The study has mainly focused on reusable construction waste. For this purpose, 7 models were developed for each type of construction waste – 5 polynomials (from the first to the fifth degree), Artificial Neural Network (ANN) and Support Vector Machine (SVM) models. The ANN models were found to be the most effective for all types of CW compared to the other models developed. The ANN models developed for CW1, CW6 and CW7 had a high R² value (> 0.85), indicating their potential for predicting the amount of these types of CW in the future. The ANN models developed for the remaining types of CW had weaker performance (R² < 0.60), but their performance could be improved in the future investigations with an increase in the amount of data on CW generation. This research underscores the importance of employing advanced data modeling techniques in addressing the challenges of construction waste management. By providing accurate predictions of CW generation, stakeholders can better strategize waste reduction, recycling, and disposal efforts, thereby contributing to the sustainable development goals of minimizing environmental impact and promoting resource efficiency in the construction sector.