Improving the heart disease diagnosis by evolutionary algorithm of PSO and Feed Forward Neural Network

The considerable growing of cardiovascular disease and its effects and complications as well as the high costs on society makes medical community seek for solutions to prevention, early identification and effective treatment with lower costs. Thus, valuable knowledge can be established by using artificial intelligence and data mining; the discovered knowledge makes improve the quality of service. Until now, different researches have been carried out in order to predict heart disease based on data mining methods such as classification and clustering methods; however, what has been less noticed is the exact diagnosis of disease with the lowest cost and time. In this paper, by using feature ranking on effective factors of disease related to Cleveland clinic database and by using Particle Swarm Optimization as well as Neural Network Feed Forward Back-Propagation, 13 effective factors reduced to 8 optimized features in terms of cost and accuracy. The assessment of selected features of classified methods also showed that PSO method along with Neural Networks of Feed Forward Back-Propagation has the best accurate criteria of the rate of 91.94% on these features.