Applying random forest to forecast municipal solid waste generation from household fuel consumption

• A random forest model was developed to forecast MSW generation • The model incorporated socioeconomic and demographic factors as independent variables • Household cooking fuel consumption data was utilized for indirect prediction • Household cooking fuel expenditures are the most significant predictor of MSW generation • The random forest model achieved a prediction accuracy of 55% Accurately forecasting municipal solid waste (MSW) generation is essential for designing efficient waste management systems and promoting sustainable urban development. As cities expand and consumption patterns shift, reliable data-driven approaches are increasingly necessary to address the complexities of MSW generation. This study applied the random forest (RF) algorithm, a machine learning technique, to predict MSW generation at the household level. RF was selected for its capacity to handle non-linear relationships, imbalanced datasets, and outliers. The analysis focused on data from year 2019, avoiding distortions associated with the COVID-19 pandemic. The model integrated per capita MSW data with household fuel consumption indicators (i.e., natural gas, electricity, and liquefied petroleum gas) and demographic variables such as age, education level, and monthly expenditure. The case study focused on the city of Lima, Peru, using 80% of the data for training and 20% for testing, with hyperparameters optimized via 5-fold cross-validation. The final model explained 55% of the variance in MSW generation (R² = 0.55). This result reflects the model’s ability to capture significant drivers of variability, although it leaves room for refinement due to factors not included in the analysis, such as cultural practices or seasonality. Among the predictors, household monthly expenditure on cooking fuels emerged as the most influential variable, reinforcing the connection between resource consumption and waste generation. These findings highlight the potential of integrating socioeconomic indicators into predictive models to enhance their reliability. By improving forecasting capabilities, this study supports targeted policies for urban waste management and sustainable resource use.