Machine Learning Techniques to the Prediction of Variables of the Urban Solid Waste Collection Process

Due to rapid urbanization, improved living standards, and climate change concerns, solid waste management has taken on a large dimension today. As a result, solid waste collection and transportation are considered one of the most critical elements of any solid waste management system, as these activities account for a significant fraction of management costs. Moreover, the concern for their efficiency has further increased with the emerging modern era, which has led the industry to make serious efforts to revolutionize the solid waste (SW) collection process, seeking sustainability and cost-effectiveness through advanced technologies and intelligent systems where Artificial Intelligence (AI) techniques have gained momentum by offering computational approaches as an alternative to solve SW management problems. Therefore, in this study, it is proposed to implement Machine Learning Techniques for the prediction of 4 critical variables in the urban solid waste (USW) collection process, such as total route time, kilometers traveled, number of compactions performed by the collection vehicles, and tons collected during the route. Eleven regression-based machine learning techniques are used through the application of 3 experiments. Results show that fully connected neural networks can predict mentioned variables with a correlation coefficient R
²
higher than 0.7.