A Machine Leaning-Based Approach To Predict The Cervical Cancer

Abstract

The leading cause of high mortality rates in the world continues to be cervical cancer. It is essential to put in place a comprehensive strategy that includes prevention, early diagnosis, screening, and treatment programs to address this. For long-term success, utilizing technology can offer a competitive advantage. By identifying risk patterns from medical records, machine learning-based predictive models have demonstrated promise in predicting patient outcomes, leading to higher survival rates through early detection. The right machine learning techniques must be used for accurate cancer diagnosis. In this thesis, we suggest the application of the Logistic Regression (LR), Random Forest (RF), Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost) algorithms and Ensemble Method (Bagging) to cervical cancer prediction. The technique Synthetic Minority Techniques (SMOTE) is used to address these problems because the dataset used in this study has missing values and is highly unbalanced. This has improved detection of cervical cancer in patients. In order to reduce the complexity of the classifier and the amount of computational work needed, feature selection is a crucial pre-processing step that is frequently used to determine the most important input characteristics. Grid search and Random Search are then used to further improve the results after that. The suggested algorithms were Ensemble method (Bagging) which combine XGBoost and LR meta-learner because they produced the better results. With an accuracy of 98.83% and an F1 score of more than 80%, the acquired findings are quite strong and encouraging. This could be seen as a sign that future iterations of these synthesis techniques should be executed in an effort to improve the accuracy of cervical cancer prediction.