Skip to main content

Table 3 Comparison between different feature selection techniques and the proposed ensemble method for k=100, on the TCGA dataset

From: Automatic discovery of 100-miRNA signature for cancer classification using ensemble feature selection

Classifier Random GALGO EFS-CLA UFS EN LASSO RFE EFS
Gradient Boosting 0.8588 0.8782 0.8871 0.9028 0.9208 0.9315 0.9309 0.9359
Random Forest 0.8515 0.8787 0.8824 0.8929 0.9224 0.9341 0.9288 0.9324
Logistic Regression 0.8015 0.8295 0.8832 0.8813 0.8988 0.8996 0.9088 0.9237
Passive Aggressive 0.6986 0.7235 0.8111 0.8091 0.8406 0.8424 0.8506 0.8831
SGD 0.7278 0.764 0.8446 0.8334 0.8649 0.8648 0.8824 0.9035
SVC 0.8077 0.8348 0.8706 0.885 0.9049 0.9008 0.9103 0.9154
Ridge 0.6534 0.6614 0.7422 0.7504 0.7753 0.7751 0.7954 0.8305
Bagging 0.822 0.8382 0.8562 0.8719 0.8889 0.9078 0.9061 0.911
Global Average 0.7777 0.8010 0.8472 0.8534 0.8771 0.8820 0.8892 0.9044
Calls to Classifier - 60,000 480 - - 10 946 80