Table 2. Comparison of 17 ileal gene signatures selected by four different feature selection methods.
Methods | AUC | Accuracy |
Boosting | 0.928 | 89.9% |
PAM | 0.895 | 88.9% |
Random forest | 0.902 | 85.9% |
LASSO | 0.895 | 85.9% |
Boosting [16], PAM) [17], random forest [18] and LASSO [19] were applied to the SAM filtered training microarray dataset to select 17 ileal gene signatures. The AUC and overall accuracy for each of the signatures were calculated based on the majority vote of 7 classifiers (Boosting, PAM, Random Forest, LASSO, Support Vector Machine, Linear Discriminant Analysis, and Naive Bayes), which is equivalently to the decision based on the median score using an usual probability threshold of 0.5 (see Materials and Methods).