The identification of aged-rice adulteration by support vector machine classification combined with characteristic wavelength variables

The aging of rice during long-term storage leads to a decline in quality and a decrease in consumption value. Therefore, the efficient and rapid detection of aged-rice adulteration is crucial for rice quality control. This paper aims to utilize near-infrared spectroscopy (NIRS) in conjunction with chemometrics to identify two types of rice adulteration: Late japonica rice and Southern japonica rice. To achieve this, competitive adaptive reweighted sampling (CARS) and successive projections algorithm (SPA) were utilized to select characteristic wavelengths from the spectral data, both with and without pretreatment methods. In order to establish a support vector machine classification (SVC) model with exceptional predictive capability, cross validation (CV), genetic algorithm (GA), and particle swarm optimization (PSO) were adopted to find the best combination of parameters (C, g). Notably, the prediction accuracy achieved by MMN-CARS-CV-SVC, SNV-CARS-CV-SVC, SNV-CARS-GA-SVC, and SNV-CARS-PSO-SVC models was 98% for Southern japonica rice adulteration, with varying proportions of aging rice. Conversely, the prediction accuracy of the MSC-CARS-CV-SVC model reached 96% for Late japonica rice adulteration, with varying proportions of aged rice. In conclusion, the combination of NIRS and SVC enhances the discriminative analysis of aged-rice adulteration, and the outstanding generalization ability of the SVC model offers new possibilities for the analysis of adulteration in other cereal grains.