A novel VOC mixtures classification methods based on GBLinear and TabNet and informative feature selection from gas sensors (E-Nose) data

The GBLinear and TabNet algorithms have been incorporated with essential feature selection techniques to create a new method of classifying essential oils using e-nose systems. Essential oils, known for their complex chemical compositions and a wide variety of applications in industries such as food, cosmetics, and pharmaceuticals, pose some challenges for e-nose systems due to the high variability and subtle differences in volatile compounds (VOCs). This novel approach, used for the first time for the analysis of electronic nose data, integrates efficient machine-learning models with advanced feature selection techniques and aims to increase the accuracy and interpretability of essential oil classification. This study highlights the potential of integrating interpretable machine learning models with deep learning-based architectures to address challenges in the analysis of complex gas mixtures. Not only was the classification accuracy increased by these methods, but these methods could be used in the future as promising models for analyzing complex mixtures.