Abstract:

In this study, 157 kiwifruits from Meixian county, Shaanxi were selected and investigated for accurate evaluation
of kiwifruit quality. Nine indices including weight, long axis, short axis, thickness, volume, color, sugar content, acidity,
and firmness were measured and analyzed by descriptive statistics and correlation analysis. Principal component analysis
was employed to build a composite score mathematical model, and then the composite scores were analyzed by K-means
cluster. Fisher discriminate analysis was applied to re-cluster fruits for evaluating the reliability of K-means cluster analysis.
Results showed that all other grading indices except weight and volume presented significant differences. The 157 samples
were classified into three clusters according to their composite scores: good, 0.10-1.39; moderate, –0.44-0.09; and bad,
–1.27-0.46. The correct discrimination rate reached 98.72% when compared with the clustering analysis, suggesting that
there is a good consistency.

Key words: kiwifruits, quality evaluation, principal component analysis, cluster analysis, discriminant analysis