PREDICTION OF PREGNANCY-RELATED DISEASES USING MACHINE LEARNING

Abstract

Pre-eclampsia can lead to serious complications for both the mother and the baby if left untreated. Decreased blood supply to the placenta can lead to low birth weight or early delivery, as well as harm to vital organs, especially the brain, kidneys, and liver. When the body cannot manufacture enough insulin to fulfill the additional needs of pregnancy, it results in Gestational Diabetes Mellitus (GDM). Pregnant women may experience higher birth weights, which raises the possibility of birth difficulties, as well as an increased risk of type 2 diabetes later in life. Anemia may raise the risk of low birth weight or premature delivery, as well as make it more difficult to recover after giving birth. In order to help pregnant women take the appropriate precautions, we attempted to predict these three disorders before the start of the third trimester. We used a number of machine learning classifiers, both individual and group, as well as an artificial neural network that had been properly pre-processed. In our dataset, we addressed the class imbalance by applying the SmoteNC approach. In order to enhance the PE prediction performance even more, we have employed boosting and bagging models. AdaBoost, the light Gradient Boosting Machine, and ANN yield the best-performing models, with corresponding auc-roc scores of 0.846, 0.911, and 0.879 for PE, GDM, and anemia. We also determined the most significant features for each of the three disease categories by using the calculated feature importances.