Hybrid Deep Learning Models For Predicting Gestational Diabetes Mellitus
DOI:
https://doi.org/10.64252/mbsqfc48Keywords:
Gestational Diabetes Mellitus (GDM), Hybrid Deep Learning, GRU (Gated Recurrent Unit), LSTM (Long Short-Term Memory), RNN (Recurrent Neural Network), SMOTE (Synthetic Minority Oversampling Technique), Pima Indian Diabetes Dataset, Machine Learning, Random Forest, XGBoost, Model Accuracy, Pregnancy Health Prediction.Abstract
Diabetes Mellitus forms part of our nation's serious public health concerns. It could be a clutter of the digestion system in thousands of the individuals emerging from a statement of extra glucose inside the body. If diabetes is not diagnosed rather soon on human creatures there will continuously be serious wellbeing complications for example cardiac conditions, renal problems, nerve complications and damage to the eyes. Gestational diabetes is a pathological increase of blood glucose during pregnancy with high risk to mother and child including preeclampsia, neonatal hypoglycemia and future complications. As a result, early interventions for better maternal and fetal outcomes benefit from the accurate prediction and early diagnosis of GDM. Current models usually underperformed in predicting GDM with adequate accuracy more so with imbalanced datasets and lack of supportive mechanisms to handle them. Diabetes mellitus among lactating women is the topic of this research and it intends to predict the same by utilizing Deep learning Algorithms viz GRU+LSTM, GRU, LSTM, RNN, and also different Machine learning algorithms namely Random Forest and XG-Boost classifiers on a real–time diabetes dataset. Considerable effort is therefore put to work-including new techniques such as SMOTE-to obtain the best solution in forecasting of diabetes mellitus. The Pima Indian Diabetes Dataset from the National Institute of Diabetes and Digestive and Kidney Diseases provided the data—all female patients, 768 in all, above the age of 21. One target variable and eight predictor variables are available. In this research, after using the SMOTE algorithm, accuracies were achieved and achieved in whose sequence were GRU+LSTM- 99%, RNN-96%, Random Forest-85%, XG-BOOST-86%, GRU-90%, LSTM-82% are better than those of rest of models covered in the research. The accuracy of existing model [3] comparing the proposed model with respect to XG Boost, (84%), Random Forest (83%), Decision Tree (79%); these proposed models gave the accuracy of GRU+LSTM-99%, RNN- 96%, XG-Boost-86%, Random Forest-85%.
