Fig. 2.

Analysis of characteristic volatile compounds of Cabernet Sauvignon wines from the Xinjiang region. a Heatmap cluster of analysis of the relationship between the winemaking area and the related volatile compounds (distance measured using Euclidean and clustering algorithm using Ward’s D). b Random forest analysis of the importance of characteristic volatile compounds of each winemaking area. The features were ranked by the mean decrease in classification accuracy when they were permuted