Discrimination of Chinese prickly ash origin place using electronic nose system and feature extraction with support vector boosting machine

Chinese prickly ash (CPA) is renowned for its distinct flavors across various regions in China. In this study, we present a novel approach using an electronic nose (E-nose) system to discriminate CPA samples originating from Henan, Gansu, Sichuan, Yunnan and an additional region. A total of 300 samples, with 60 samples from each of the five regions, were tested. The corresponding signals of the E-nose were initially fitted using polynomial regression methods, followed by the application of convolution methods to extract features from the polynomial parameters. These novel techniques were coupled with a support vector boosting machine for origin place classification. The hyperparameters of the model were optimized using the Harris Hawk optimization algorithm. Comparative analysis was conducted with the t-distributed stochastic neighbor embedding method and principal component analysis. Additionally, the proposed model was benchmarked against eight state-of-the-art methodologies. Empirical results demonstrate the superior performance of the model, achieving an impressive accuracy of 95.17% when retaining five features per sensor. This work offers a valuable and innovative approach to accurately discriminate the origin place of spices.