Similarity model by matching and cross entropy-driven methods to support tracing source of unknown waste

Abstract

There is a growing recognition that illegal dumping or global transfer of solid waste poses an environmental challenge. The dearth of effective tracing source techniques exacerbates the difficulty in the identification of unknown wastes, thus further complicating the environmental and management challenges. To this end, we developed tracing source processes for unknown waste, leveraging similarity models to facilitate identification. With a dataset of waste features, we established a matching function for single waste feature, as well as a cross entropy model for multiple waste features. Both the similarity models were applied to the tracing source process, enabling the identification of the source or category of unknown waste. The similar probability value between known waste and unknown waste can be obtained by those two models. The process of source tracing in the study was shown by examples of aluminum dross. If the known waste feature dataset is sufficiently accurate, the accuracy rate of tracing source will be correspondingly high in practical applications. Therefore, when using the similarity models, it is imperative to improve the known waste dataset to satisfy the demands of actual tracing source.