Multistage SVM as a Clinical Decision Making Tool for Predicting Post Operative Patient Status.
Melissa Stockman, Mariette Awad · 2010
Abstract- Because applying machine learning techniques in support of clinical decision would improve decision makers in healthcare, we present in this paper a comparative framework of Support Vector Machine (SVM) classifiers based on post operative patient (POP) data. We compare the performance of a single multiclass SVM and a multistage SVM (MSVM) to those obtained by a number of other classifiers presented in the literature and show that both SVM approaches significantly outperform the other methods resulting in 84.4 % and 94.3 % overall accuracy respectively. Results for the non-SVM classifiers ranged from 48 %- 77.7 % accuracy.