Prediction of heart disease using hybrid technique for selecting features

Generally Healthcare industry is known to be `information rich', but woefully all the data required to discover hidden patterns are not mined. For effective decision making in field of medical, advanced techniques of data mining are used. This paper proposed a prediction of heart disease using random forest and naive bayes. In addition, approach is proposed to select features before classification in order to improve performance of models. For feature selection, SVM-RFE and gain ratio algorithms are applied to dataset which in results assigns weight to each feature. This approach helps to improve accuracy and reduce computational time. Experimental results shows that proposed approach of selecting feature increases accuracy for both models.

I. Introduction

Heart disease is result of alteration of functionality and structure of heart. This leads to inadequate pumping due to which organs and tissues receives insufficient amount of oxygen for their metabolic needs. Tissues could become necrotized due to lack of oxygen and may cause death. Heart problem is one of the major reason of death in Australia, USA, Canada, UK. Threateningly all doctors do not possess competence in every field and moreover there is shortage of medical expertise at certain places. Heart disease is said to be imperative reason of fatality in the number of different countries. In US said disease shatters 1 individual in every 34 second.