Abstract
Early diagnosis of arrhythmia is very significant because the symptoms are not always prominent and may go unnoticed until a deteriorating heart health condition is reached. Due to progress in machine learning techniques and the availability of a larger dataset, it is now possible to analyze and propose better techniques to detect cardiac arrhythmia. This paper presents a new hybrid approach that uses feature selection followed by ensemble classification for the diagnosis of arrhythmia. The feature selection algorithm provides a list of strongly correlated features with arrhythmia and at the same time making sure to use minimum redundant information. Using the arrhythmia dataset from UCI for training and testing purposes, the performance of the proposed hybrid approach had 77.27% accuracy and 76% precision. This outperformed the eight individual estimators used in this paper for comparison.

This publication has 28 references indexed in Scilit: