New approach for forecasting demolition waste generation using chi-squared automatic interaction detection (CHAID) method

Abstract

The purpose of this study is to propose a classification method for building demolition waste (DW) that is different from existing studies and to develop a demolition waste generation rate (DWGR) prediction model. To achieve the purpose, the chi-squared automatic interaction detection (CHAID), which is a decision tree (DT) method, was used in this study. Additionally, 796 buildings were measured using the data collection method that calculated the quantity of waste generation through actual measurements immediately before the building removal. The results using CHAID allows us to easily understand the complex influencing relationships between the DW types and various factors influencing the DW generation. Furthermore, the CHAID method was developed for forecasting the DWGR. Then, split-sample validation was performed to confirm the prediction performance of the CHAID algorithm applied in this study. The results show that the prediction performance of the current study is higher than that of the previous studies. In particular, the CHAID model for concrete classifies approximately 98.9% of the concrete generation correctly. Because the CHAID model of this study was developed from previous building cases, it can assist construction companies and building demolition contractors in decision making.