Optimized Stacking Ensemble Classifier for Early Cancer Detection Using Biomarker Data

Ovarian cancer ranks sixth globally as a major cause of death among women, with a five-year survival rate below 50%, largely due to late detection. Early detection is crucial to lower mortality rates. This paper introduces an Optimized Stacking Ensemble Classifier (OSEC) for early ovarian cancer detection using biomarkers. The model comprises two layers: the first layer includes base classifiers optimized with Particle Swarm Optimization (PSO), while the second layer is a meta-classifier integrating Support Vector Machine (SVM), Logistic Regression (LR), and Random Forest(RF) models fine-tuned through grid search. Among the three datasets evaluated, the Blood Routine dataset showed the best performance with a stacked RF meta-classifier, achieving: 94.29% accuracy. The Stacked RF model also outperformed others, reaching 92.82% accuracy on the Serum dataset and 92.77% on the Malignant Ovarian Tumor (MOT) dataset, consistently excelling in precision, recall, and f1-score.