Comparison on Confidence Bands of Decision Boundary between SVM and Logistic Regression
Xing Wang, Xin Wang, Zhaonan Sun · 2009
Support Vector Machine (SVM) and Logistic Regression (LR) are two popular classification models. The main purpose of a classification algorithm is to figure out the estimator for the decision boundary. In this paper, we considered confidence bands of decision boundary generated from SVM and LR. Confidence bands of decision boundary are estimated through bootstrap methods. We compared the confidence band estimator of SVM with the estimator of the conventional LR. Our main result is that sample size of the observations makes effect on the stability of both SVM and LR, sample size ratio, central location and covariance matrix of the data bring less effects on the stability of SVM than that of LR.