Abstract: The dielectric loss factor ?’’ spectrum and relative dielectric constant ?’ spectrum of jujube fruits from three different varieties, Lingwu Changzao, Dongzao, and Tuanzao, were measured with an LCR meter at 55 selected frequency points in the frequency range of 1–1 000 kHz. Effective information from the dielectric spectra was extracted by principal component analysis (PCA) and genetic algorithm (GA). Models for variety identification of jujube fruits were established using partial least squares discriminant analysis (PLS-DA), linear discriminant analysis (LDA) and support vector machine (SVM), respectively. The results indicated that both the frequency and varieties had a significant influence on the dielectric parameters of jujube fruits. The model built using the effective information extracted by PCA and GA was better than that established the original spectra, and the SVM model was better than the PLS-DA and LDA models. The PCA-SVM model based on dielectric loss factor ?’’ was better than the GA-SVM model based on relative dielectric constant ?’, with the former model giving 100% correct discrimination for prediction set. Therefore, the PCA-SVM model based on dielectric loss factor ?’’ spectrum was optimal for variety identification of jujube fruits.

Key words: jujube fruit, dielectric spectrum, variety, identification model