Modified Rotation Forest Ensemble Classifier for Medical Diagnosis in Decision Support Systems

Decision support system (DSS) in medical diagnosis helps medical practitioners in assessing disease risks. The machine learning algorithms prove a better accuracy in predicting and diagnosing diseases. In this study, rotation forest algorithm is being used to analyse the performance of the classifiers in medical diagnosis. The study shows that rotation forest ensemble algorithm with random forest as base classifier outperformed random forest algorithm. In this study, we use linear discriminant analysis (LDA) in place of PCA for feature projection in modified rotation forest ensemble method for classification. The experimental result also reveals that LDA can provide better performance with rotation forest while comparing with PCA. The accuracies given by random forest, rotation forest and proposed modified rotation forest classifiers are 89%, 93% and 95%, respectively.