Municipal Solid Waste Forecasting in China Based on Machine Learning Models

With the rapid development of industrialization and urbanization, urban sustainability has been challenged by ever-worsening environmental problems and calls for a smart strategy to monitor their dynamics across different regions and countries. In China, municipal solid waste (MSW) accounts for a large proportion of the overall urban waste, which characterizes China as the largest MSW producer in the world. Except for the huge population, an alternatively possible reason is the lower utilization efficiency than other developed countries. In this paper, we use provincial panel data spanning from 2008 to 2019 in China and discuss the construction of a proper MSW prediction model. Based on the machine learning technology, we mainly focus on six models in comparison, such as MLR, SVR, RF, KNN, XGBoost, and DNN. We select nine explanatory variables highly related to regional development and industrial structure, as well as MSW generating characteristics, and explored the prediction performance of each model (with the best parameters and structure) under three different preprocessing strategies. By introducing the SHAP value method, we proceed to identify the correlation between explanatory variables, explained variables, and important explanatory variables. Our findings confirm that all machine learning models performed high prediction accuracy, among which the DNN model was the best under different preprocessing strategies. In addition, the regional GDP indicator play the most important role in predicting the MSW output. Moreover, the increase of urban population and agglomeration of wholesales and retails industries can positively promote the production of MSW in regions of high economic development, however, in less developed regions the performance could be opposite.