HGSORF

HGSORF predicts cesarean section (C-section) deliveries using a Henry Gas Solubility Optimization-tuned Random Forest and provides explainable analysis of contributing factors.

Key Features:

• Optimization Algorithm: Henry Gas Solubility Optimization (HGSO) is used to fine-tune Random Forest hyperparameters and help avoid local minima during optimization.
• Predictive Accuracy: Reported performance includes 98.33% accuracy on the Pakistan Demographic and Health Survey (PDHS) dataset.
• Model Comparison: Performance is compared against Gaussian Naive Bayes (GNB), Linear Discriminant Analysis (LDA), K-nearest Neighbors (KNN), Gradient Boosting Classifier (GBC), and Logistic Regression (LR).
• Data Balancing: ADAptive SYNthetic (ADASYN) algorithm is applied to balance class distributions in the dataset.
• Explainability: eXplainable AI methods SHAP (SHapley Additive exPlanation) and LIME (Local Interpretable Model-Agnostic Explanations) provide local and global explanations of model predictions.

Scientific Applications:

• Clinical Decision Support: Provide predictions and explanatory factors to inform clinical decision-making regarding cesarean deliveries.
• Research and Development: Support obstetrics research by evaluating predictive algorithms and feature contributions on demographic health survey data such as PDHS.

Methodology:

Random Forest hyperparameters are optimized using HGSO; ADASYN is used to balance the dataset; model performance is evaluated on the PDHS dataset (98.33% reported) and compared to GNB, LDA, KNN, GBC, and LR; SHAP and LIME are used for local and global explainability; HGSO performance is compared against traditional hyperparameter-optimization algorithms.