Abstract: 【Objective】 To address the issues of low efficiency in traditional grape variety identification methods and poor interpretability of machine learning models,providing a scientific basis for rapid field identification of grape varieties.【Methods】 An efficient identification model for wine grape varieties was constructed based on leaf morphological traits. Utilizing the support vector machine-recursive feature elimination（SVM-RFE） method,seven major wine grape varieties in Yantai,Shandong Province,were selected as research objects,for which seventeen leaf morphological traits were measured. The Bayesian optimization（BO） algorithm was used to optimize the parameters of the support vector machine（SVM）,K-nearest neighbors（KNN）,and decision tree（DT） models. Model performance was evaluated based on AUC value,and key features were screened using RFE combined with the corresponding model classification accuracy.【Results】 The SVM model optimized by BO exhibited the best performance,achieving an AUC value of 0.960 7,a precision of 95.56%,a recall of 84.31%,and an accuracy of 95.46%,with an F1 score of 0.895 8 when setting Cabernet Gernischt as the positive class. RFE screened out 14 key features,and the constructed BO-SVM-RFE model maintained high performance while achieving an improved model accuracy of 96.05% and further enhancing model interpretability.【Conclusion】 A varietal identification model for wine grapes based on SVM-RFE（the BO-SVM-RFE model） was successfully constructed,and 14 key identification indicators,including petiole length,were clarified.

Key words: Wine grapes, Varietal identification, Support vector machine, Recursive feature elimination, Leaf morphological traits, Bayesian optimization