Ensemble methods for heart disease prediction

This work consists of automatic heart disease prediction ensemble methods; this critical human health task is performed using several new algorithms. First, we introduce a weak classifier based on the randomness analysis of binary sequences. Second, we present another classifier in which the shrunk covariance estimation is utilised during the training and prediction phases. Third, we present a classifier in which Gaussian probabilities are summed via a kurtosis and KS-test importance scheme. Finally, a two-fold ensemble implementation is created by fusing logistic regression and our majority voting density estimation classifier. This final classifier is compared with state-of-the-art methods, and the sensitivity, specificity, accuracy and optimised precision are reported.