Fig. 1
From: Hellinger distance-based stable sparse feature selection for high-dimensional class-imbalanced data
Decision score of the SVM classifier varies greatly as the class-imbalanced ratio changes. In this simulation, the X-matrix (1000×3) is randomly produced then fixed while the class label y=±1 is randomly changed at each iteration under the fixed class-imbalanced ratio. The prediction point is set be x0=(−5,9,−1)T. The black line and the red vertical short segments are, respectively, the mean and standard deviation of prediction decision score with the class-imbalanced ratio changing from 1 to 99 based on 2000 iterations