Accurate and efficient identification of chemical compounds is essential in various scientific and industrial applications. Vibrational spectroscopy techniques such as Fourier Transform Infrared Spectroscopy (FTIR) is a ubiquitous analytical technique that has been proven to be highly valuable method in chemical analysis as well as surface and materials characterization . In the context of high-throughput experiments, automatic data analysis methods are becoming essential, since manual data analysis is time -consuming process prone to human errors.

This study presents a machine learning approach that leverages FTIR data for the accurate identification of pure chemical compounds. We introduce a modular and interpretable ML framework that integrates classical chemometric preprocessing with robust supervised learning algorithms for FTIR spectral classification. Four classifiers—PLS-DA, XGBoost, Random Forest, and SVM—are evaluated on a chemically diverse set of pure compounds. Our tailored data extraction process effectively captures the major peaks characteristic of each compound, thereby enhancing the model's efficiency. Our results demonstrate that the used approach not only improves the identification accuracy but also reduces computational complexity, making it a robust tool for rapid and exact compound identification. Additionally, a researcher-friendly desktop application with an integrated GUI is presented to streamline the full pipeline.

This methodology holds significant potential for applications in pharmaceuticals, environmental monitoring, and chemical manufacturing, where swift and reliable compound analysis is paramount.