Fetal state classification from cardiotocography based on feature extraction using hybrid K-Means and support vector machine

Abstract

Cardiotocography (CTG) records fetal heart rate (FHR) signal and intra uterine pressure (IUP) simultaneously. CTG are widely used for diagnosing and evaluates pregnancy and fetus condition until before delivery. The high dimension of CTG data are the problem for classification computation, by extracting feature we can get the useful information from CTG data, and in this research, K-Means Algorithm are used. After extracting useful information, data are trained by using Support Vector Machine (SVM) to obtain classifier for classifying the new incoming CTG data. Based on 10 cross validation, this method have a good accuracy to 90.64% using Cardiotocography Dataset obtained from UCI Machine Learning Repository. Data are classified into fetal state normal, suspicious, or pathologic class based on seven abstract features that extracted from twenty one original features and then trained using hybrid K-SVM Algorithm. This research shows the ability and capability of Hybrid K-SVM for classifying CTG dataset. In general, the experimental result of hybrid K-SVM show the better classification compare to SVM.