Construction and Demolition Waste Generation Forecasting Using a Hybrid Intelligent Method

Abstract

To effectively forecast the construction and demolition waste (C&DW) generation, this paper proposes a hybrid intelligent method, which integrates singular spectrum analysis (SSA), support vector regression (SVR) and long and short-term memory (LSTM) network. The key idea of this method is to use SSA to divide the original time series into trend and fluctuation sub-series, separately perform LSTM and SVR approach, then to combine the forecasts generated from each model. Grid search optimization is further utilized to seek for the optimal parameters of the proposed forecasting model. Experiments are conducted to compare the proposed method and several benchmark approaches. Results show that it is statistically significant that the proposed method can provide much better forecasts than other approaches, indicating that our method can be used as a promising tool for the C&DW generation forecasting.