Application of Feature Selection Algorithms to Select Elements that Distinguish Regional Differences in Chinese Grown Winter Jujube Fruit ( Zizyphus jujuba Mill. cv. Huanghua Dongzao)

Abstract

XGBoost, machine-learning algorithms, was applied for the first time to select the minimum number of elements needed to distinguish Huanghua winter jujube fruit (from Hebei Province) from winter jujube grown in other regions of China. Toward this end, ICP/MS and AAS were used to measure the concentration of 30 elements in fruit and soil. XGBoost screened the thirty elements, selected ten of them, and then differentiated Huanghua fruit from the other regions studied with an overall authenticity accuracy of 95%. Compared to logistic regression and support vector machine algorithms, XGBoost was more efficient, producing higher accuracy rates, improving farmer and consumer protection against food fraud.