Model structures and approaches for municipal solid waste prediction – A review

Understanding the generation patterns of municipal solid waste and adapting to it is one of the most urgent problems of the decade, which requires meaningful answers in order to establish sustainable and resilient waste management infrastructures. The purpose of this literature review is to systematize the models and approaches that are suitable for predicting the amount of municipal solid waste (MSW). Based on systematically explored literature analysis (used Scopus database in 2023) methodologies, a recommendation framework is developed to underpin the design of municipal solid waste management systems by recording possible models and proxy variables in addition to presenting their advantages and limitations based on the 78 papers that met the selection criteria. The proposed framework identifies possible model structures according to the municipal solid waste management analysis task and data availability, which fills the gap and lays the foundations for data-based decision support in goal-oriented municipal solid waste management. A total of 36 AI-driven models are explored and catalogued using 455 proxy variables. Depending on the number of observations and the set of available proxy variables, a model structure framework is provided, based on which the identification of models can be optimized for the completion of goal-oriented municipal solid waste predicting tasks. This literature review provides a basis for meeting the goal-oriented MSW modeling tasks at a higher level, as it also proposes a model catalog, data inventory and model structures, and discusses how to address the challenges that may arise for the models. It is the first research to distinguish modeling issues of interest to industry professionals and academics, thereby providing guiding results for the entire municipal solid waste management sector. • 36 models used to forecast the amount of municipal solid waste are explored. • A model structure recommendation framework for MSW forecasting is developed. • The model selection depends on the set of variables and the number of observations. • Neural networks are the most widely used methods for waste amount forecasting. • Model accuracy can be improved by increasing the input data or by external variables.