Fig. 1From: Fecal source identification using random forestDesign of the classifiers. (a) Determination of the bacterial group profiles for each of the source samples. (b) Analysis of the bacterial profile discrepancies between the samples that belong to the targeted source and all other samples using random forest (one classifier per source). (c) Selection of the relevant representative sequences based on mean decrease Gini values to form the classifier. Classifiers are then trained using the selected representative sequences. Abbreviations: MDG mean decrease GiniBack to article page