The correlations among the rankings of features based on different RF parameters (Figure 6) were very high and significant (r > 0.9), suggesting that each of the RF parameters would rank the features in a very similar order while classifying the individuals into either the AUD or CTL group. High correlations among these parameters also suggest that each parameter is very valuable and reliable in terms of its classification performance, lending further support to the utility of the RF technique as a powerful tool for classifying individuals using a set of multi-domain features.
