Integrating Potentiostat Measurements and Ensemble Learning for Water Pollution Estimation

Microplastic pollution in water has emerged as a serious environmental concern due to its persistence and impact on aquatic ecosystems. Detecting microplastics in aqueous environments remains a complex task, particularly across different concentration levels and due to the lack of observable visual indicators. This study presents a machine learning approach for estimating microplastic concentrations using current-voltage signals generated by a potentiostat. These signals were processed through a feature extraction stage that identified six numerical descriptors, including peak current, voltage, and area from upper and lower signal regions. The resulting feature set was used as input for several machine learning algorithms, including Random Forest, K-Nearest Neighbors, Support Vector Machine, Logistic Regression, and an ensemble learning. Model evaluation was conducted using stratified 5-fold cross-validation to ensure balanced data partitioning. Performance was further enhanced through hyperparameter optimization using GridSearch. Among all tested models, the ensemble learning achieved the best results, with an accuracy of ${9 2. 3 3 \%}$ after optimization, outperforming individual models in all evaluation metrics. These findings support the potential of ensemble learning strategies in improving the reliability of microplastic estimation based on potentiostat signals and offer a foundation for more scalable monitoring tools in future water quality assessment systems.