Optimizing municipal solid waste collection management through data mining: a case study in southern Brazil

Abstract

This study presents three models based on urban solid waste collection data from three municipalities in southern Brazil to identify collection patterns. With the support of Knowledge Discovery in Databases and Data Mining techniques and algorithms, historical data on the weight of unloaded waste from collection trucks in transfer stations, collection route data, and socio-demographic and climate data were used to predict the amount of solid waste collected at each point and assess collection patterns. Data were collected, pre-processed, modeled, and analyzed using Linear Regression, Gradient Boosting, and Random Forest algorithms. Our results show that the Gradient Boosting algorithm model performed better: Mean Absolute Error (25.244), Root Mean Square Error (87.667), and Coefficient of Determination (0.642). In this sense, this study contributes in two ways: first, it helps organizational decision-making and improves the collection service provided to the local community. Second, this study collaborates with the scholarly literature reinforcing the potential of data mining for urban solid waste management.