Hybrid Machine Learning Approach for Early Detection of Ovarian Cancer using SVM and KNN Models with Feature Selection

Early and accurate detection of ovarian cancer remains a significant challenge in clinical practice. This study proposes a hybrid machine learning model combining Support Vector Machine (SVM) and K-Nearest Neighbors (KNN) to enhance the detection of ovarian cancer by integrating biomarkers and physiological features. We systematically evaluate feature selection techniques, including Recursive Feature Elimination (RFE) and statistical methods, to identify the most relevant predictors. The study uses biomarkers (e.g., CA-125, HE4) and physiological data (e.g., age, BMI) to create a multi-modal dataset. The features are tested individually and in combination to assess their impact on model performance. The hybrid model integrates predictions from SVM and KNN using majority voting, leveraging the strengths of both algorithms in global and local classification. The model is evaluated using various metrics, including accuracy, precision, recall, F1-score, and AUC, demonstrating improved detection accuracy. Results show that incorporating both types of features significantly enhances predictive performance, underscoring the potential of multi-modal data integration for developing robust diagnostic tools for ovarian cancer.