Prediction of gestational diabetes diagnosis using SVM and J48 classifier model

Abstract

Knowledge Discovery in Databases (KDD) process is also known as data mining. It is a most powerful tool for medical diagnosis. Due to hormonal changes, diabetes may  occur during pregnancy is referred as Gestational diabetes mellitus (GDM). Pregnant Women with GDM are at highest risk of future diabetes, especially type-2 diabetes. This paper focuses on designing an automated system for diagnosing gestational diabetes using hybrid classifiers as well as predicting the highest risk factors of getting Type 2 diabetes after delivery. One of the common   predictive data mining tasks is classification. It classifies the data and builds a model based on the test data values and attributes to produce the new classified data. For detecting GDM and also its risk factors, two classifier models namely modified SVM and modified J48 classifier models are proposed. The data set were collected from various hospitals and clinical labs and preprocessed with discretize filter using weka tool. Missing values are replaced by the suitable values. The final preprocessed data applied in the proposed classifier Model.  The output of the proposed model is compared with all the other existing methodologies. Since the proposed model modified J48 classifier model produces more accuracy and low error rate against other existing classifier models.