An evolutionary machine learning approach for municipal solid waste generation estimation utilizing socioeconomic components

Municipal solid waste generation is an important parameter in waste management with significant impacts on environment. There are many components directly influencing solid waste generation (SWG). In this study, the important socioeconomic parameters of the city of Tehran, Iran, were obtained in the period of 1991 to 2013. Important and optimum variables were analyzed and selected using the Pearson correlation analysis, and four variables including income, pop, GDP, and month were selected. In order to find a proper model, three common machine learning (ML) techniques including artificial neural network (ANN), random forest (RF), and multivariate adaptive regression splines (MARS) were used. Five evaluation metrics were used in this study including the correlation coefficient, Nash coefficient, root mean square error (RMSE), mean absolute error (MAE), and ratio of RMSE to the standard deviation of measured data (RSD). The results revealed that the MARS model outperformed all the other models. In addition, in the last step, the crow search algorithm (CSA) was applied to the MARS model to increase the accuracy of the selected model. The hybrid optimized MARS-CSA model resulted in better prediction, where the correlation coefficient increased to 0.82, the Nash coefficient decreased to 0.56, and the errors decreased to RMSE = 10.908, MAE = 9.141, and RSD = 0.657. In practice, for better understanding and efficient management of municipal SWG and reducing the environmental impact of waste, an integrated MARS-CSA model can be suggested for the accurate prediction of the monthly SWG as an essential pre-requirement.