Prediction of medical waste generation using SVR, GM (1,1) and ARIMA models: a case study for megacity Istanbul

Abstract

Purpose Estimation of the amount of waste to be generated in the coming years is critical for the evaluation of existing waste treatment service capacities. This study was conducted to evaluate the performance of various mathematical modeling methods to forecast medical waste generation of Istanbul, the largest city in Turkey. Methods Autoregressive Integrated Moving Average (ARIMA), Support Vector Regression (SVR), Grey Modeling (1,1) and Linear Regression (LR) analysis were used to estimate annual medical waste generation from 2018 to 2023. A 23-year data from 1995 to 2017 provided from the Istanbul Metropolitan Municipality's affiliated environmental company ISTAC Company were utilized to examine the forecasting accuracy of methods. Different performance measures such as mean absolute deviation (MAD), mean absolute percentage error (MAPE), root mean square error (RMSE) and coefficient of determination (R-2) were used to evaluate the performance of these models. Results ARIMA (0,1,2) model with the lowest RMSE (763.6852), MAD (588.4712), and MAPE (11.7595) values and the highest R-2(0.9888) value showed a superior prediction performance compared to SVR, Grey Modeling (1,1), and LR analysis. The results obtained from the models indicated that the total amount of annual medical waste to be generated will increase from about 26,400 tons in 2017 to 35,600 tons in 2023. Conclusions ARIMA (0,1,2) model developed in this study can help decision-makers to take better measures and develop policies regarding waste management practices in the future.