• P-ISSN 0974-6846 E-ISSN 0974-5645

Indian Journal of Science and Technology


Indian Journal of Science and Technology

Year: 2019, Volume: 12, Issue: 2, Pages: 1-7

Original Article

Artificial Neural Network versus Binary Logistic Regression for Determination of Risk Factors of Myocardial Infarction


Objectives: To identify the important (significant) risk factors of Myocardial Infarction (MI) and construction of statistical models using conventional technique of binary Logistic Regression (LR) and of artificial Neural Network (NN). Both the statistical models (LR vs NN) are compared in their predictive capabilities. A case-control study with the purpose of comparison of LR outcomes to NN outcomes. The research is covering the whole country. Therefore, the required data is collected from all parts of Pakistan (Peshawar, Quetta, Karachi, Lahore, Islamabad etc). The required data is collected in 13 months; starting from 01-Feb-2013 to 30-Mar-2014. Materials and Methods: The research is basically a case-control study. For this purpose a sufficient sample size of 2,000 is included containing 1,000 patients (cases) and 1,000 controls. The samples are collected from various places of the country. The sample involves male and female. AMOS and SPSS are used in the study to analyze the collected data. Two techniques are applied to the data to identify the significant risk factors of MI i.e. LR and Artificial NN. Results obtained from LR and NN are compared. Findings: Out of total 28 potential risk factors of MI, 16 variables are found significantly associated to the MI by LR analysis and 17 (16 are those selected by LR) are found significantly associated to MI by ANN model. Only one variable differs between the two outputs, i.e. fried food intake. The rest of 16 variables are exactly the same. These 16 risk factors are: hypertensive disorder, higher age, family history of CVDs, chest pain, atherosclerosis, psychosocial pressure, alcohol use, diabetes mellitus, income class, breathing problem, smoking, fish intake, obesity, male gender, physical activity, vegetable intake, and often intense anger. All the 16 risk factors are significant in development of the disease. In this study the most threatening etiology is found to be chest pain. Applications: The outcomes of the study has drawn the attention of the epidemiological investigators to consider other procedure (NN) alongside the orthodox method (LR) while examining risk factors of myocardial infarction for a better insight and comparison purpose. The results show that all the clinical and modifiable risk factors are important in context of Pakistan.

Keywords: Logistic Regression, Myocardial Infarction, Neural Network, Risk Factors, AMOS, SPSS


Subscribe now for latest articles and news.