Identification of binary mixture composition of carbohydrates based on terahertz multi-parameter spectroscopy and machine learning

To ensure the quality and safety of high-end carbohydrates, this study proposes a method for identifying the ratio of carbohydrate binary mixtures based on terahertz spectroscopy multi-parameter fusion and machine learning. Multi-parameter spectral data (including absorption, refractive index, and transmission spectra) of carbohydrate samples were collected using a transmission terahertz time-domain spectroscopy (THz-TDS) system; three machine learning algorithms were combined to construct binary mixture ratio identification models, and the optimal scheme was determined. Owing to the non-destructiveness and high penetrability of terahertz spectroscopy, it can accurately capture the intrinsic optical differences of carbohydrate molecules with different ratios in the terahertz range. This method reduces human errors, shortens analysis time, rapidly and accurately identifies the ratio of carbohydrate binary mixtures in complex matrices, and has important practical application value in component verification in fields such as food and health products.