Multifunctional Electrodes for Signal Soil Measurements: Benchmark With ML-Based Algorithms

Soil monitoring is a crucial issue for sustainable field and agricultural management. This article explores the performance of machine learning models in classifying soil types under varying moisture levels using wireplate and plateplate sensor configurations. Voltammetric sensors can be used to analyze soil in situ and act as a microbial fuel cell (MFC). Several classifiers were applied, and the logistic regression and support vector machine achieved the highest accuracy, reaching 99% in the wireaplate configuration and 94% in the plateaplate configuration. Focusing on soil classification under different moisture conditions, Adaboost and random forest outperformed other models, achieving an accuracy of 93% and 90%, respectively. The study highlights the importance of sensor design, model selection, and environmental factors in optimizing soil classification accuracy. These findings suggest that tailored machine learning approaches, in combination with refined sensor configurations, can improve the reliability of soil monitoring systems in agricultural and environmental applications. Furthermore, the integration of MFCs enables simultaneous soil characterization and energy harvesting, enhancing the potential for self-sustaining monitoring solutions.