Prediction of congenital heart defects in children using data mining techniques

Abstract

Introduction: Congenital heart abnormalities are one of the most important global health problems and the most common type of birth abnormalities. Many factors can contribute to the development of these disease. Congenital heart defects can occure as a result of a combination of genes with other risk factors such as exposure to environmental factors, maternal nutrition or drug use. Predicting congenital heart disease based on related risk factors can have an important role in preventing patients such abnormalities and improving preventive strategies in this regard. Materials and Methods: Type of this research is an applied cross-sectional study which has been conducted on 791 cases of patient records. Data for this research has been collected from medical files between 2007 and 2017. The database of this research includes 31 features. In this research, data mining algorithms were implemented on the data set to build a prediction model. The attributes were weighted and ranked through using the feature selection algorithm. Results: According to the results, the decision tree model was the best model for predicting congenital heart anomalies with 98.6% accuracy. Based on the SVM feature selection algorithm, the attributes of consanguineous marriage and hydraminosis were determined as the most effective features in predicting congenital heart abnormalities. Conclusion: The proposed model based on mining algorithms has the capability to predict congenital heart defects based on maternal risk factors during and before pregnancy. Keywords: Data Mining, Congenital heart defects, Machine Learning, Prediction Model