Ovarian cancer is one of the most challenging gynecological cancers to diagnose early, which significantly impacts survival rates. This study presents a comparative analysis of machine learning (ML) and deep learning (DL) models for the early prediction of ovarian cancer using clinical and biomarker data. The dataset undergoes comprehensive preprocessing, including handling missing values, outlier removal, normalization, and dimensionality reduction via SelectKBest. Feature selection methods such as Feature Importance, and autoencoder-based techniques are employed to enhance model performance. A variety of classifiers, including LightGBM, CatBoost, and a hybrid model (Random Forest + Gradient Boosting), along with deep learning models like Multi-layer Perceptron (MLP), are evaluated. Ensemble models, including the Voting Classifier, are also implemented. The results demonstrate that the CatBoost model, combined with optimized feature selection techniques, achieved the highest accuracy of 88%, showcasing its potential as a reliable tool for the early prediction of ovarian cancer. This study emphasizes the importance of integrating optimized preprocessing, feature engineering, and appropriate model selection to achieve effective early diagnosis.

Keywords: Ovarian cancer, early prediction, machine learning, deep learning, feature selection, CatBoost, LightGBM, hybrid model, ensemble learning, predictive modeling, clinical data, biomarker data, Recursive Feature Elimination, SelectKBest, Multi-layer Perceptron, Voting Classifier.