Analysis of Resampling Method for Arrhythmia Classification Using Random Forest Classifier with Selected Features

There are many developed techniques to be experienced for biomedical data mining. Due to the growing number of cardiac patients, many data are stored by different organizations. It helps to work with data mining for medical signals. Here In this work, an attempt for analysis and classification of arrhythmia cardiac disease is taken by authors. Data collected from the UCI repository to validate the purposed method. The features from the data are chosen using correlation-based feature selection (CFS) method due to the elimination of redundant data. For better processing and classification the selected data is to be resampled and is done in this work using random sampling. Random forest classifier is used for classification purpose. The performance is evaluated with resampled data and exhibited in result section. It is found that the average accuracy is 96% with random resampling technique.

I. Introduction

Machine learning based classification techniques can be used for decision-making in the healthcare domain which includes disease diagnosis and prediction, psychoanalysis, survival analysis, and hospital management. In recent years many countries in the world also adopting an automated diagnosis system to detect a disease at an early stage. This automatic system can detect a disease by analyzing the data collected from different medical test [1]. Machine learning based diagnosis system is a supportive tool in the clinical environment where the machine is trained with different features and class of a disease. The class of a disease refers to the decision of a physician after analyzing several features of a disease.