关键词: 非浓缩还原梨汁；评价体系；因子分析；聚类分析；判别分析

Abstract: To explore the relationship among the quality indexes of not from concentrate (NFC) pear juice and construct a comprehensive evaluation system, 32 pear cultivars were selected to investigate 12 quality indexes by conventional descriptive statistics, correlation analysis, factor analysis and regression analysis, and then a discriminant model for predicting NFC pear juice quality was established by K-means cluster and discriminant analysis. The results showed that the dispersion degree varied greatly among the 12 quality indexes with variation coefficients ranging from 5.46% to 105.73%. The flavonoid content had the largest variation coefficient of 105.73%, while the color parameter L value had the smallest variation coefficient of 5.46%. Four common factors were extracted from the converted data matrix by factor analysis, including functional factor, flavor factor, appearance factor, and sweetness factor, respectively, contributing to 27.873%, 24.890%, 17.364% and 14.235% (84.362% together) of the total variance. By regression analysis, the total phenol content, sugar/acid ratio, hue angle (h value), soluble sugar content, L value and flavonoid content were selected as the evaluation indexes for NFC pear juice quality to establish discriminant functions for the grading of NFC pear juice quality with a recognition accuracy of 100% for modeling samples. The discriminatin functions could be applied for discriminate the comprehensive quality of NFC pear juice.

Key words: not from concentrate pear juice; evaluation system; factor analysis; cluster analysis; discriminant analysis