Ensemble Based Foetal State Diagnosis

Since decades, cardiotocogram recordings have been used worldwide in detecting foetal abnormalities. Computational methods are found to be more effective than manual interpretation of the signals. This paper deals with the classification of cardiotocogram data using artificial intelligence methods to help doctors in decision making. Based on the interpretation of the cardiotocogram signal, the doctor decides whether to carryout normal delivery or operation. The data for analysis that is obtained from the UCI machine learning repository consists of information extracted from the cardiotocogram signals corresponding to normal and pathological foetal state. Machine learning methods such as naive Bayes, random forest, J48 and a stacking model were used to classify the foetal state. Highest balanced accuracy of 98.6% was attained by the proposed stacking model, which consisted of naive Bayes as meta-learner and naive Bayes, random forest and J48 as base learners. Accurate prediction of pathological foetal state has a crucial role in ensuring the foetal well-being. When compared to the individual classifiers and other algorithms the stacking model predicted pathological foetal state most accurately with a success rate of 98.9%.