MACHINE LEARNING FOR CLASSIFICATION OF SPECTRUMS INFRARED OF HIGH SIMILARITY BIOLOGICAL SAMPLES

• Grant number: 25/03317-2
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: August 01, 2025
• End date: July 31, 2026
• Field of knowledge: Physical Sciences and Mathematics - Physics
• Principal Investigator: Leandro José Raniero
• Grantee: Bianca Siqueira Massaroto
• Host Institution: Instituto de Pesquisa e Desenvolvimento (IP&D). Universidade do Vale do Paraíba (UNIVAP). São José dos Campos , SP, Brazil

Abstract

Fourier Transform Infrared (FTIR) Spectroscopy is widely recognized for providing non destructive analyses capable of identifying the chemical structure of different materials, and it is applied across various fields of science due to its versatility. However, the complexity of the generated data often requires more advanced analytical tools, motivating the use of machine learning algorithms to detect relevant patterns and features. In this sense, three supervised classification models are adopted-Support Vector Machine (SVM), Random Forest, and Logistic Regression-which stand out for their ability to handle high volumes of information and for consistently performing well in classification tasks. To evaluate the performance of each model in specific contexts, metrics such as accuracy and precision are employed, ensuring an objective and well-founded comparative analysis. It is expected to understand which machine learning approach best adapts to the different needs in interpreting FTIR data and contributes to advances in both academic research and industrial applications. In this study, two different groups of samples will be employed, one composed of saliva and the other of blood serum. The saliva group will be formed by normal children and children diagnosed with autism spectrum disorder, while the serum group will be composed of normal adults and diabetics. Thus, it is intended to verify the percentage of correct classifications by the models on highly similar data. (AU)