Waste Generation Forecasting Using Mobility Data During COVID-19 Pandemic

Municipal solid waste (MSW) generation forecasting serves as the basis for future waste management strategic plans. However, rapid socio-economic and environmental changes lead to fluctuations in the MSW generation patterns. This is particularly exacerbated during unforeseen events, such as the COVID-19 pandemic. This study investigated the effect of COVID-19 and mobility conditions on the behavior of MSW generation predictors. Multiple artificial intelligence models were employed, including generalized linear models decision trees (DTs), support vector machines, deep learning, random forests (RFs), and gradient-boosted trees (GBTs). The models were tested around a 5-year historical MSW generation, incorporated with various socio-economic, seasonal, and climatic factors, in Abu Dhabi, United Arab Emirates. Multiple data split ratios were examined: 1) 60/40, 70/30, and 49/51 to simulate the COVID-19 pandemic. The feature importance analysis indicated that socioeconomic had the highest impact on generation rates, with a Pearson correlation coefficient of 0.30-0.71, followed by seasonal and climatic factors, with 0.12-0.40 and 0.15, respectively. The forecasting results indicated that training the models with 70% of the dataset resulted in higher accuracy compared to using a 60% training split. GBTs and RFs demonstrated superior performance, with root mean squared error (RMSE) of 0.029 and 0.031, respectively. On the other hand, DT achieved the lowest accuracy of 0.057 RMSE. The analysis revealed a significant improvement in the models' predictive capabilities upon the inclusion of COVID-19 and mobility conditions by 17-41%. The findings of this study emphatically highlight the importance of incorporating contextual variables into MSW generation prediction methodologies, affirming their significant impact on enhancing forecasting accuracy in rapidly changing environments.