Meta-heuristic Swarm Intelligence based algorithm for feature selection and prediction of Arrhythmia
The healthcare industry is developing extensively, thus more high-dimensional datasets are emerging. Thus it is imperative for health maintenance industries to mine healthcare data effectively, that can be used for accurate prognosis of diseases. Since, all the available features in these datasets are not valuable and more number of irrelevant features may negatively affect the performance of classifier model. Therefore, accuracy of models can be augmented by using a prudently selected subset of features, instead of using each available feature. Feature subsets yield superior result as compared to entire set of features. However, choosing effectual and key features is a tedious task in healthcare datasets. Therefore, in this paper, an algorithm based on Ant Colony Optimization and Simulated Annealing concept is proposed for selecting valuable feature subset in high-dimensional arrhythmia dataset for detection of disease using decision tree. As arrhythmia is considered to be a grievous disease, so timely detection and prevention would be valuable for patients. Experimental results indicate that the proposed algorithm outperforms other approaches in terms of accuracy.