Using machine learning, this study predicts waste composition by county to improve recycling and waste-to-energy planning. It also builds the first open dataset to help governments and educate the public on better waste habits.
According to the World Bank’s “What a Waste 2.0” report, municipal solid waste (MSW) is projected to reach 3.4 billion metric tons by 2050, representing a 70% increase from current levels. Factors like population growth, rapid urbanization, and economic development are considered to be the main contributors to this projected rise. To promote and establish effective waste reduction, recycling, and reuse strategies, it is critical for governments and municipalities to identify the types and quantities of materials in waste streams. We develop a machine learning model to predict MSW material composition at the county level using socio economic factors such as population density, urbanization rate, and employment sectors. For this purpose, we also collect and harmonize data from municipal waste audits across the U.S., creating the first publicly available dataset of its kind. We utilize SHAP (Shapley Additive exPlanations) value plots to explain the output of ML model and importance of each feature. We evaluate the model performance using the coefficient of determination. A random forest model with 24 selected features achieves a high level of accuracy over the test set. The results of this research could contribute to the data-driven planning of MSF infrastructure such as recycling facilities and waste-to-energy plants. Furthermore, data on waste composition could help guide educational campaigns to encourage waste segregation and promote sustainable behaviors.
Have trouble accessing an IISE presentation? Login to your account at IISE Conference Archives and click "Search Archives"
If you are interested in the Sustainable Development division of the Institute of Industrial and Systems Engineers (IISE), become a member today!