Mojtaba Heravi - Master of Science student, Department of Knowledge Engineering and Decision Sciences, University of Economic Sciences, Tehran; Member of Young Researchers and Elite Club, Islamic Azad University, Qazvin Branch
Saeed Setayeshi - Associate Professor, Department of Medical Radiation Engineering, School of physics and nuclear engineering, Amirkabir University of Technology, Tehran;

Diseases had been the greatest threat for human being along the history. Heart diseases (HD) have gained special attention in medical studies. Recently studying about intelligent method such as classification and diagnosis of HD has been continuing as a key topic. Points of view in researches have been done in order to increase precise and reduce error in this kind of decisions. Aim of this paper is to propose a simple hybrid-linear model using logistic regression and single layer perceptron neural network which is using after a special pre-processing, identification of noisy data. And so, based on Pareto’s law, the network training with only 20% of the data exist is performed. The model for improving the classification and patterns recognition of HD has been used on clinical data of 270patients from the Cleveland Clinic (UCI website).The model has been implemented in MATLAB. The mean-error of the proposed model on the total dataset was 11/11%, which was achieved a significant improvement compared to recent similar methods. The results clearly show that the linear proposed technique has more effects on reducing the error in the classification and identification of patient more accurately and short-time than conventional methods and complex nonlinear. The method can help a doctor for early detection of disease or as a decision support system (DSS).

Heart diseases diagnosis, Biomedical engineering, Classification, Pattern recognition, Machine learning, Artificial neural network, Single layer perceptron, Logical regression, Pareto’s law