Development of an Predictive Model via Decision Tree-Based Algorithms for Forecasting Demolition Waste Generation

Management of demolition waste (DW), which accounts for a large portion of waste generation (WG), is a very important issue. Therefore, many researchers tried to apply various ML algorithms to predict WG, and tried to find the decisive factors affecting WG. This study conducted a study on the development of optimal ML model for predicting demolition waste generation (DWG). In this study, decision tree (DT), random forest (RF), and gradient boost machine (GBM) algorithms were applied to develop ML models to predictive DWG. For this, data preprocessing was performed and the optimal hyper parameter was searched for each algorithm to derive an optimal ML model. In consideration of dataset size, leave one out cross validation (LOOCV) was applied to the model validation and mean absolute error (MAE), root mean square error (RMSE), coefficient of determination (R squared), and mean square error (MSE) were used as the performance evaluation index of the models. As a result of this study, it was found that the predictive performance of the RF model (MAE 72.837, MSE 12198.236, RMSE 110.446, R² 0.880) was better than one of DT (MAE 87.081, MSE 17348.052, RMSE 131.712, R² 0.829) and GBM (MAE 87.883, MSE 18175.125, RMSE 134.815, R² 0.821) models. The error from the observed mean (987.1806 kg m^-2) was 8.82%, 7.38%, and 8.90% for the DT, RF, and GBM models, respectively. Therefore, it can be seen that the ML model using the DT-based algorithms is very good at predicting DWG. Finally, this study presented a reliable and optimal ML model for predicting DWG for a domestic waste management strategy.