Integrating Machine Learning and Cox Models for Enhanced Survival Prediction in Ovarian Cancer

Ovarian cancer remains one of the deadliest gynecologic malignancies due to late-stage diagnosis and complex disease progression. Accurate survival prediction is vital for improving personalized care and clinical decision-making. This study proposes a hybrid survival analysis framework combining machine learning (ML)-based feature selection with the Cox proportional hazards model to enhance survival prediction. Feature importance is determined using LASSO and Random Forest algorithms, and these selected variables are used in the Cox model to compute hazard ratios and survival probabilities. The proposed framework achieves superior performance (C-index 0.82, IBS 0.143) compared to traditional and ML-only models, as validated by Kaplan-Meier curves and log-rank tests. This hybrid approach balances predictive accuracy with clinical interpretability, demonstrating its potential in data-driven oncology.