Ovarian Cancer Diagnosis: A Data-Driven Machine Learning Approach

Ovarian cancer is a highly fatal gynecological disease due to late-stage diagnosis and non-specific early symptoms. This paper presents a machine learning pipeline using clinical data from 3,000 patients, incorporating imputation, encoding, normalization, SMOTE for class balancing, RFE for feature selection and five classifiers. Among them, the Random Forest classifier achieved the highest overall accuracy at 88.94%. This indicates that almost 90% of the model’s predictions came true, indicating that it is highly generalisable across various classes within the ovarian cancer risk spectrum. The findings emphasize the effectiveness of ensemble models and feature engineering in early detection. Models were evaluated using accuracy, precision, recall, F1-score, ROC-AUC and confusion matrices with hyperparameter tuning via GridSearchCV and an 80/20 stratified split. The approach addresses class imbalance and improves model interpretability.This work demonstrates the potential of ML pipelines in clinical decision support for ovarian cancer diagnosis.