Machine Learning-based Prediction of pH and Temperature using Macromodel of Si3N4-gated Transistor

Abstract

Machine learning algorithms are employed in sensing applications for data processing and analysis, such as extracting different features and predicting specific parameter. This work predicts discrete pH levels and temperatures using decision tree and neural network algorithms. The input dataset was obtained from the I-V characteristics of the LTspice-simulated macromodel of the Si3N4-gated transistor-based pH sensor. Different types of decision tree and neural network models were trained and investigated using the classification learner app in MATLAB©. The performance of the ML algorithms was evaluated based on their accuracy, scatter plots, and confusion matrices. The wide neural network predicted correct pH levels with an accuracy of 99.1% against 71.9% of the fine decision tree algorithms.