Analyzing uncertainty in cardiotocogram data for the prediction of fetal risks based on machine learning techniques using rough set

The key focus of this venture is to evaluate the calibration of classifiers built on rules, trees, and functions by exploring the uncertain information that exists in the Cardiotocography (CTG) dataset. Classification is imperative in diagnosing the health of the foetus and new born specifically in critical cases. It facilitates the obstetricians in acquiring the information of foetal well-being in pregnancy, substantially for the woman with complications. The research aims to classify the CTG data points into normal, suspicious and pathologic. Rules, trees, and function-based classifiers are applied in machine learning for predicting the health of the new born. Particle Swarm Optimization (PSO) is used in pre-processing for selecting the relevant features. Rough set approximations are exploited in extracting the uncertain information from the data set. The result reveals the importance of useful information present in the uncertain data during classification. In this paper, the overall highest accuracy is displayed by Random Forest classifier with 99.57% and a tree-based approach has shown its supremacy over other approaches.