This study addresses the current lack of a robust evaluation system for Xanthoceras sorbifolium Bunge germplasm. Using 60 germplasms from Xinjiang, we conducted a comprehensive assessment of 19 indicators encompassing fruit morphology, seed phenotype, and oil/fatty acid composition. A rapid screening model was constructed by integrating probability grading, entropy weight, and Random Forest algorithms. Methodologically, correlation analysis, cluster analysis, and principal component analysis (PCA) were first employed to reveal intrinsic trait relationships. PCA indicated that PC1 (43.6% variance) primarily reflected fruit morphology, while PC2 (17.5%) was associated with fatty acids. Subsequently, based on a normal distribution, seven indicators that passed the test were subjected to probability-based grading (categorized into Grades I-III). The entropy weight method was used to determine indicator weights (e.g., the weight for seed diameter reached 17.11%). Finally, a random forest model was developed for classification evaluation. Trait variation analysis showed that single fruit shell weight (ShW) and single plant yield (Yld) had the highest coefficients of variation (reaching 36.43% and 29.09%, respectively). Correlation network analysis revealed a significant negative correlation between nervonic acid and oleic acid, but positive correlations with eicosenoic acid and erucic acid. The comprehensive scores derived from the entropy weight-probability grading method showed high consistency with the PCA results. The random forest model achieved an accuracy of 0.9412, with feature importance analysis identifying single seed weight (SeW) as the dominant factor. This research establishes a practical framework and an accurate tool to accelerate the selection and breeding of superior X. sorbifolium varieties. • elucidating the variation in 19 important indicators encompassing fruit morphology, seed phenotype, and the composition and content of oils and fatty acids. • performing a systematic comprehensive evaluation of the 60 germplasm accessions using correlation analysis, cluster analysis, and principal component analysis. • classifying certain indicators of X.sorbifolium based on probability grading. • By combining the entropy weight method and the RF, a rapid screening model for X. sorbifolium germplasm was constructed.
Zhou et al. (Fri,) studied this question.