A Data-Driven Method for Constructing the Geography of Waste in Singapore

A substantial amount of microplastics are derived from the fragmentation of plastic waste. Robust planning of the waste management system requires reliable information on waste generation. It is an essential step to forecast the temporal-spatial variations of waste in the region for a given timespan. However, no study has covered the forecasting of spatial distribution of the subcategories of municipal solid waste, which are essential when designing the specific waste-to-resource/energy pathways for a circular economy. In the case of micro-plastic waste, it is important to know how the microplastic waste geographical distributes in order to better control it from source or remove it from the environment. To fill the research gap, this project focuses on developing a data-driven method for the prediction of future waste production and mapping the geography of waste for different waste categories. A “Geography of Waste” (GoW) approach containing the steps of data collection and analysis, visualization, data-driven modelling, and waste prediction is proposed in this work. It is applicable to any type of wastes including microplastics. With the retrieved dataset, different machine learning algorithms including neural network and gradient boosting are employed to identify how significance each factor affects the waste generation compared with linear interpolation. Results show that the gradient boosting model produces a reasonable prediction on the test dataset. In future work, the model will be applied to study the geography of microplastics, and the prediction models will be integrated with the decision support tools for the planning of waste management facilities.