Figure 9
From: Classifying transcription factor targets and discovering relevant biological features
Feature Elimination. YIR018W (Yap5) was used as a prototype to determine the number of features needed to build a classifier. This graph shows the accuracy of classifiers for YIR018W built using different numbers of features selected using SVM-RFE. The x-axis gives the number of features selected for each tested classifier. Error-bars show one standard deviation of 50 classifiers constructed at each feature number using different sub-sampling of the negative set. 1500 features were selected as the point where accuracy reaches a plateau.