The t-test selection method gave similar prediction rate.
In this paper, we propose three test-selection methods and two coverage identification metrics.
Moreover, the following two subsections further demonstrate such an exceptional performance is impossible from overfitting because our proposed algorithm works well consistently for different data sets with different training and test data selection methods.
Table 3 The bootstrap based classification accuracies, stability index and number of reselected genes in all bootstrap samples of the SVM-classifier and the LDA-classifier based on all tested gene selection methods, on the SRBCT dataset aThe number of selected genes is set to 30.
However, among the tested feature selection methods, only the Ridge regression model performed satisfactorily in the generalizability assessment; therefore, it was selected as the preferred method for survival-signature modeling.
Table 1 The bootstrap based classification accuracies, stability index and number of reselected genes in all bootstrap samples of the SVM-classifier and the LDA-classifier based on all tested gene selection methods, on the LUNG dataset.
Figure 7 Stability index s2(bar chart) and accuracy of classification (dot chart) with the 95% confidence interval of the best classifier on the tested feature selection methods for SRBCT data.
The number of reselected genes is the sum of non zero bootstrap-based feature ranked genes (BBFR) Figure 6 Stability index s 2 (bar chart) and accuracy of classification (dot chart) with the 95% confidence interval of the best classifier on the tested feature selection methods for MLL data.
In order to reduce the number of OpCode n-gram features, which ranges from thousands to millions, we used the DF measure to select the top 1,000 features and tested three feature selection methods.
We tested five feature selection methods: QER, CHI2, IG, DFD, and OCFS on both Turkish and English review datasets.
Future work will include developing similar models for TFs from other structural classes and organisms, as well as refining the feature selection procedure and testing other feature selection methods (Maldonado and Weber, 2010; Nguyen and de la Torre, 2010; Yang and Ong, 2010) that might help us identify sequence features relevant for model accuracy.
