Fig. 2From: An ensemble-based feature selection framework to select risk factors of childhood obesity for policy decision makingThe main procedures of BFSMR. There are mainly four steps, (1) splitting, (2) mapping which draws bootstrapped samples from each chunk, (3) reducing which merges sets with the same Set ID and applies one feature selection classifier on each set and derive one selected feature list, and (4) merge procedure which combines the selected feature lists into the final outputBack to article page