Multidimensional analysis of parameters performed in DB10F.

Comparison of classifiers using a multidimensional analysis of classifiers evaluated in DB10F. -value represents the percentage of cases where the random configuration of parameters yields a classification that outperforms the classification obtained with default parameters. Mean, deviation and maximum refer to the increase in accuracy provided by the random configuration, when it outperforms the default configuration. In the case of the SVM, the random choice of parameters yields a classification more accurate than the default classification in about 99% of the cases. When this scenario occurs, the average and maximum improvement of quality are 20% and 39%, respectively.