Decision support in prediction of metabolic syndrome with data mining methods

Abstract

Introduction:The aim of this study was to find the most important risk factors which have a role in causing metabolic syndrome and also to evaluate the efficacy of different models by data mining.Material and Methods:We used the data of third phase of “Isfahan Healthy Heart Program” as data set, which was done on 9572 subjects in 2007. In this study, we evaluated the efficacy of 3 main algorithms including decision tree, Naïve Bayes and neural network to detect the subjects with metabolic syndrome. Results:The results of the study showed that BMI is the most significant factor leading to metabolic syndrome. The other risk factors included LDL-Cholesterol, age, education, type of employment, sex, physical activity, history of diabetes, hypertension, stroke, income, smoking history of hyperlipidemia, myocardial infarction, and heart rate. Conclusion:We found that the optimal algorithm might be different by the dataset and data preprocessing methods. Various factors have a role in the efficacy of algorithms; using data preprocessing methods increased the prediction accuracy of all the examined techniques.Our results showed that artificial Neural Networks model has the highest accuracy in the pattern recognition while Naive Bayes was better in predicting the metabolic syndrome among healthy subjects.

Keywords

• Data Mining
• Decision Tree
• Bayes Theorem
• Neural Networks
• Cardiovascular Disease