Estudio Comparativo de Métodos de Aprendizaje Automático para el Modelado de Adsorción Basada en Grafeno en el...

Polynomial regression (PR) is one of the most commonly used approaches to predict pollutant removal efficiency, but its flexibility has not been well studied. Machine learning techniques such as support vector machines (SVM) and artificial deep learning neural networks (ANNDL) have been applied in adsorption studies, but the method optimization for small-sized datasets remains limited. In this study, we evaluated four published datasets (graphene-related nanomaterials) with relatively small data sizes (20-30) using PR, SVM, and ANNDL to predict water pollutants removal efficiency. Our results indicated that although PR demonstrated reliability under cross-validation, its performance could be highly affected by data size and parameter limitations. In contrast, the radial basis function kernel SVM and ANNDL with the tensor flow framework showed higher performance and tolerance in data size and parameters than PR. This study provided recommendations to streamline experimental design and develop robust models for optimizing adsorption systems in future studies on water pollutant treatment.