Machine Learning Models for Predicting Prolonged Labor: A Statistical and Comparative Analysis using Augmented Clinical Data

Prolonged labor is a major obstetric complication that has implications for both maternal and newborn. This research seeks to undertake a clinical analysis of data from St. Philomena's Hospital of Bengaluru, India, to identify factors associated with extended second-stage labor and evaluate predictive models for its classification. The objective of this study is to predict prolonged labor using different Machine Learning methods methods. The dataset contains 280 records, which were enhanced to capture numerous maternal and neonatal characteristics, such as parity, gestational age, delivery mode, and Apgar scores. The data preprocessing steps included oversampling using SMOTE as a technique for handling class imbalance and encoding categorical data. Five machine learning models including support Vector Machine, Random Forest, XGBoost, AdaBoost, and Artificial Neural network, were built and tested based on accuracy, precision, recall and F1-score. The maximum accuracy was noted as 96.43% by SVM. This research highlights that machine learning has great potential in producing better predictive models that may be helpful in early detection of labor results and the improvement of clinical decision-making to provide a better care for mothers and healthier newborns.