Machine Learning-Based Classification of VOCs for Breath Analyzer Linked to Different Chronic Diseases

This study compares machine learning frameworks for detecting and classifying gas mixtures of ammonia, acetone, and formaldehyde using L-Ascorbic acid-reduced graphene oxide (LAA-rGO) in a Field Effect Transistor (FET) sensor optimized by gate voltage. As real-time gas monitoring becomes essential in environmental, health and industrial applications. This research addresses the identification of volatile organic compounds (VOCs) linked to chronic kidney disease, lung cancer, and diabetes. The framework leverages the enhanced properties of LAA-rGO to improve sensor sensitivity and selectivity, employing algorithms like Support Vector Machines (SVM), Naïve Bayes (NB), and Decision Tree (DT). The DT demonstrating the highest accuracy. Experimental validation confirms the framework's effectiveness in differentiating similar gases overcoming challenges like cross-sensitivity and non-linearity. Results indicate that integrating LAA-rGO with machine learning significantly boosts detection accuracy and response times, paving the way for future applications in real-time monitoring and emphasizing the benefits of combining advanced materials with data analysis for enhanced health safety and quality control.