Fetal health classification by support vector machine based on feature vector optimization

In recent years, diagnosis of fetal health has prevented fetal hypoxia at birth and reduced infant mortality. The acquired data from Cardiotocography (CTG) signals are used to monitor and evaluate the status of the fetus during pregnancy and delivery. Visual analysis of CTG signals using common guidelines is difficult and time-consuming due to the physiological complexity of the fetus. Artificial intelligencebased computer-aided diagnosis (CAD) systems can be useful in objective medical decision-making and digital health. In this paper, an optimized SVM classification method is proposed, which was used to classify UCI fetal health data. The optimization process was based on the genetic algorithm and the FDR relationship, in which the importance of all the features was obtained in the form of coefficients. By applying these coefficients to the feature vector of the data set, the separation of data into different classes is improved. First, the support vector machine (SVM) algorithm was used without optimization for data classification, and the correct classification rate (CCR) was 8 8. 2 8. Then, by normalizing, whitening and removing outlier data in the pre-processing stage, the accuracy increased to 9 0. 3 0. By applying the feature value coefficients to the input of the class band model with the aim of optimizing the feature vector, the classification accuracy increased to 93.98.