Statistical and Machine Learning-Based Predictive Models for Gestational Diabetes Mellitus Prevention

The focus of this paper is to use machine learning to create predictive models that detect the probable factors impacting Gestational Diabetes Mellitus (GDM) which is developed in some pregnant women. GDM is defined as any proportion of glucose intolerance developed during pregnancy. Several factors may cause GDM complications. Here, we aimed to identify factors predisposing to GDM and predict the occurrence based on several predictive models. The dataset used in this study is the Pima Indian. With the assistance of Machine Learning and Statistical Analysis, it is possible to develop intelligent models that are capable of making decisions on an autonomous basis. Seven machine learning models were tested to determine which model fits the dataset better. These models learn from past instances of data through Statistical Analysis and pattern matching. Based on the learned data, they provide us with the predicted results. This study establishes the feasibility of machine learning in the field of public health. It is observed that each technique gives different results of associated factors. The Cascade classifier model attained an accuracy of 98.58%, Random Forest (89%), SVM (69%), Logistic Regression (78%), K-NN (72%), and Decision Tree (78%). These models are validated and evaluated using several metrics. This work demonstrated that identifying risk factors must not consider one model.