An Enhanced Method for Diagnosis of Bacterial Vaginosis based on Support Vector Machines with Linear Kernel

Abstract

Bacterial Vaginosis (BV) is a pathological condition that causes complications in women’s health. Efforts to characterize it have failed to reveal a BV etiology. In this work, the Support Vector Machine (SVM) is used as base classifier in three different scenarios to identify between classes of VB. The first scenario uses the entire feature set in the dataset. The second scenario uses two sub-datasets created with the features in two rankings obtained from previous work. The third scenario uses one feature at a time to create classifiers. Performance measures in each are given. The dataset used is a real vaginal microbiology test of 201 women from Tabasco, Mexico. Results show SVM surprisingly obtained 100% accuracy in a classifier made of a single feature. This research is a first effort to lay the groundwork for computer-based BV diagnosis as advice.