Deep Learning Model for Gestational Diabetes Prediction Based on Imbalanced Data and Feature Selection Optimization

Gestational diabetes mellitus (GDM) is a form of elevated blood sugar which appears in pregnancy. It can happen at any point during pregnancy as well as create problems both for the mother and the child, both before and after birth. Giving GDM patients an early and accurate diagnosis is essential for effective treatment along with disease control as well as achieving the third sustainable development goal. Undiagnosed diabetes can lead to several dangerous conditions such as heart attack and kidney disease. This necessitates the need for learning model improvement in GDM detection and evaluation. Digital health has gained significant traction in recent years with the aim of enhancing care for diabetic pregnant women. This technology has generated an enormous amount of data that could be used to improve the management of this chronic disease. Benefiting from this, artificial intelligence (AI) methods, particularly deep learning (DL) which is a newly developed branch of machine learning (ML), are commonly used and showing good outcomes. In this paper, a Multilayer Perceptron (MLP) model is proposed to determine whether a woman has GDM. Pregnant women with and without diabetes are represented in the dataset under consideration. Imbalanced data of 1012 patients with six major features and a target column with result diabetic or non-diabetic have been analyzed and preprocessed. Feature selection optimization is developed to enhance the performance of the proposed model. The results show that the proposed model significantly improved the prediction accuracy over other related works with promising prediction accuracy which reaches almost to 98%.