Design and Development of an Ensemble Feature Selection and Machine Learning Based Cardiac Disease Prediction System

Throughout the past several decades, cardiac illnesses have been the leading cause fatalities worldwide and have become the condition that poses the greatest threat to survival. An intelligent computer-aided early prediction system may be a good alternative than relying on labor and cost intensive manual cardiac disease diagnosis. Therefore, the purpose of this research is to develop an interactive system which employs ensemble feature selection and machine learning based techniques to predict cardiac diseases at the beginning stage. To attain the objective, relevant patient dataset has been retrieved, analysed, visualized and pre-precessed. The most significant features from the dataset has been extracted using a majority voting ensemble feature selection technique that aggregates the results from six individual techniques. Following that, ten different types of machine learning classification models are employed to make the prediction based on the dataset with both all and reduced feature set. As per the performances from the machine learning classifiers, a stacked ensemble machine learning model is then proposed, trained and tested using all and reduced feature set. Five distinct versions of the proposed model are generated based on the meta-learner model preference, and their performances are assessed. The result shows that, the proposed stacked ensemble model with the reduced feature set outperforms all other classifiers with 91.7% accuracy using this particular dataset. Finally, an interactive website prototype for heart disease prediction has been developed with the reduced feature set chosen from the ensemble technique and the highest performing proposed predictive classifier.