Bio-inspired Algorithms for Diagnosis of Heart Disease

Abstract

Around one in every four deaths occur due to heart diseases in the USA every year (610,000 people approximately). One of the main reasons of fatality is due to a heart disease which also depends on various factors like obesity, diabetes, and aging. The deaths due to heart disease reduced by an indicative 41% in the USA between 1990 and 2016, whereas in our India it increased by approximately 34% from 155 to 209. The aim of this study is to aid the diagnosis of heart disease using bio-inspired algorithms. In this paper, a novel approach for the diagnosis of heart disease is inspected with the use of bio-inspired algorithms on Statlog (Heart) database from the UCI database. Bio-inspired algorithms used were binary ant colony optimization (ACO), binary firefly algorithm (FA), binary particle swarm optimization (PSO), and binary artificial bee colony (ABC) for feature selection. Bio-inspired algorithms target to decrease the dimensions of the dataset by defining the attributes which are most discerning. This helps us to successfully and efficiently classify whether a person is suffering from any heart disease or not. Out of the four algorithms, using the binary particle swarm optimization we have got the maximum accuracy of 90.09% and the classifier used was decision tree classifier. The results show that the algorithm is adequately quick and definite to be used in the analysis.