Miniaturization based Fetal Health Classification using Cardiotocographic Data

With the current growth of technology towards medicine, various ultrasound methods are available to find the fetal health. It is accessed with various clinical parameters with 2-D imaging and other test. However, health prediction of fetal heart still remains an open issue due to spontaneous activities of the fetus, the minor heart size and deficiency of knowledge in fetal echocardiography. The machine learning techniques can find out the classes of fetal heart rate which can be used for earlier forecasting. With this overview, we have used Cardiotocographic Fetal heart rate dataset extracted from UCI Machine Learning Repository for predicting the fetal heart rate health classes. The classification of fetal heart rate are achieved in four ways. Firstly, the data set is preprocessed with Feature Scaling and Missing Values. Secondly, raw data set is fitted to all the classifiers and the performance is analyzed. Thirdly, the raw data set is subjected to dimensionality reduction using PCA with 10 components and then fitted to all the classifiers and the performance is analyzed. Fourth, data set is subjected to variants of PCA like Kernel PCA, Incremental PCA, Sparse PCA and Minibatch Sparse PCA with 10 components and then fitted to all the classifiers and the performance is analyzed. Fifth, performance analysis is done using metrics like Precision, Recall, Accuracy and F-score before and after feature scaling. The implementation is done using python language under Spyder platform with Anaconda Navigator. Experimental results shows that Kernel SVM classifier is found to retain the accuracy of 98.5% in PCA, Kernel PCA, Incremental PCA, Sparse PCA and Minibatch Sparse PCA after feature scaling.t