Figure 4
From: An integrated pipeline for prediction of Clostridioides difficile infection
Feature importance for the cohort with or without (simulated) genetic data was plotted for two selected models (gbm and xgbDART), which outperformed other models (glm and nnet). This study was based on 12 features, including one genetic risk factor, rs2227036, from IL8. Feature importance from glm and nnet was always plotted as a control to compare the rank of the features weighted by optimal modeling algorithms (gbm and xgbDART) in MyCode (top two rows) and nonMyCode samples (bottom two rows). The genetic feature was weighted the top tier in gbm and xgbDART but not in glm and nnet irrespective of PSM in the MyCode cohort.
