Study of Risk Factors in Early and Late Onset Preeclampsia Prediction Using Machine Learning Classifiers

Maternal health is not only linked to a woman but also her kids and the family. Children must have healthy mothers in order for them to be healthy. Hypertension is the most common medical problem encountered in pregnancy. Incidence is seen around $\mathbf{5 - 1 0 \%}$ of all pregnancies. Preeclampsia is the second common cause of maternal mortality in India and a major cause of perinatal mortality and morbidity. The study offers a prediction model based on 103 records of pregnant women's Maternal Characteristic: AGE, BMI, Gestational Age, Systolic Blood-Pressure and Diastolic Blood-Pressure. Past History: Parity, Gravida and Past Histroy. Tree-based classifier: Decision Tree, Random Forest and Boosting classifier: Gradient Boosting, AdaBoost, XGBoost, LightGBM and CatBoost used to create the pregnancy complications prediction model. In this cohort, Majority of the mothers were diagnosed with late onset preeclampsia LOPE after 34 weeks of gestation is 70.87 %, while the remaining were having early onset preeclampsia EOPE is 29.12 %. BMI consistently emerged as the most predictive feature for EOPE and LOPE. AGE, SBP, and DBP also showed significant contributions, emphasizing the role of maternal cardiovascular and demographic factors. AdaBoost is the most balanced EOPE model, whereas RF is suitable when high recall is needed e.g., screening. LightGBM is the best model to correctly identify most LOPE cases with minimal trade-off in precision.