Figure 7.

Variable importance of the models. (a–c) Variable importance measured in the linear discriminant analysis (a), Naïve Bayesian model (b) and stochastic gradient boost model (c). The importance of the most important feature was scaled to 100.