Improving Prediction of Cervical Cancer Using KNN Imputed SMOTE Features and Multi-Model Ensemble Learning Approach.

Area of Science:

• Oncology
• Medical Imaging
• Machine Learning

Background:

• Cervical cancer is a leading cause of female mortality in developing nations.
• Early detection and treatment are crucial for minimizing adverse outcomes.
• Pap smear image analysis is a key method for identifying cervical cancer.

Purpose of the Study:

• To develop an automated system for cervical cancer prediction.
• To address challenges posed by missing values and class imbalance in datasets.
• To enhance the accuracy of machine learning models for cervical cancer detection.

Main Methods:

• Utilized a stacked ensemble voting classifier model.
• Incorporated KNN Imputer for handling missing data.
• Employed SMOTE (Synthetic Minority Over-sampling Technique) for feature up-sampling.

Main Results:

• Achieved 99.99% accuracy, precision, recall, and F1 score using KNN imputed SMOTE features.
• Demonstrated superior performance compared to models with removed missing values or only imputation/SMOTE.
• Validated the proposed model against existing state-of-the-art methods.

Conclusions:

• The developed system effectively handles missing values and class imbalance in cervical cancer detection data.
• The findings can assist medical practitioners in timely diagnosis and enhanced patient management.
• This automated approach holds potential for improving cervical cancer screening and care.