Quality evaluation of the processing suitability of yam using near-infrared spectroscopy combined with enhanced CARS-SVR algorithm

Yam possesses significant nutritional and medicinal properties that have positioned it as an important topic in functional food research. Developing scientific processing evaluation systems and detection technologies for yam's internal composition are essential for enhancing standardization and market competitiveness of yam-based products. This study was conducted in three sequential phases: initially, the optimal yam variety for steaming processing was identified from four candidate cultivars (Wencheng yam, Crispy yam, Hemp yam, and Iron Stick yam) through comparative evaluation. Subsequently, the correlations between steaming processing properties and compositional characteristics of raw materials were systematically analyzed. Finally, non-destructive detection technology was employed to enable rapid quantification of yam compositional components, thus establishing a technical foundation for standardized yam product development. Among the four yam varieties, Wencheng yam was determined to be the most suitable for steaming processing through sensory evaluation and analytic hierarchy process (AHP). Near-infrared spectroscopy combined with an enhanced Competitive Adaptive Reweighted Sampling-Support Vector Regression (CARS-SVR) algorithm enabled precise detection of four physicochemical indicators. The CARS-SVR model demonstrated superior performance for individual components, achieving R p ²= 0.921 (RMSEP=4.344) for amylopectin and R P ²= 0.900 (RMSEP=0.924) for amylose detection. This study offers an efficient detection technology for yam industry modernization.