| Keywords |
Waste management; Demolition waste; Machine learning; Predictive model; K-Nearest Neighbors |
| Abstract |
Due to the rapid increase in waste generation, smart waste management technology has become important in recent years. In this study, an optimal demolition waste generation rate (DWGR) prediction model was developed by applying various distance metrics of the K-Nearest Neighbors (KNN) algorithm. The optimal K value and the prediction model were determined through mean square error (MSE), mean absolute error (MAE), coefficient of determination (R2), and coefficient of variation of the root mean square error (CVRMSE) for Euclidean, Manhattan, and Chebyshev as KNN metrics. As a result of this study, it was found that the Manhattan-KNN (k=5) model (R2 value of 0.789) had better predictive performance than the Euclidean-KNN (k=6) model (R2 value of 0.685) and the Chebyshev-KNN (k=12) model (R2 value of 0.627) in predicting DWGR. And the mean of the observed values was 987.181 kg·m-2, and the mean of the predictive values of the Manhattan, Euclidean, and Chebyshev models were 992.307 kg·m-2, 993.144 kg·m-2 and 994.050 kg·m-2, respectively. |