ENHANCED STACKING ENSEMBLE MODEL FOR CERVICAL CANCER PREDICTION

Abstract

Early prediction of cervical cancer is crucial for positive outcome of treatments and reducing high mortality rate among women. This research aims at enhancing ensemble Machine Learning model to improve existing methods so as to have early diagnosis of the disease. Traditional methods have limitations in terms of sensitivity and specificity hence, there is need for more accurate and reliable predictive models. Existing machine learning models often face challenges such as overfitting, underfitting, and variability in prediction performance, this research which integrates different Machine Learning techniques in improving prediction, offers a potential solution to these challenges. In this study, an ensemble approach was used to develop an enhanced ensemble Machine Learning model using Support Vector Machine (SVM), Extreme Gradient Boosting (XGB) and Random Forest (RF). The three algorithms where chosen for their robustness against overfitting and improving accuracy of prediction. Furthermore the study Overcome issues of imbalanced datasets by applying Synthetic Minority Oversampling Technique (SMOTE) and Random Under Sampling (RUS) before training and testing the Machine Learning algorithms so as to obtain better result and improve accuracy of the model. This study achieved a higher accuracy of 95%, 97% precision, recall of 96% and 97% F1-score. The result indicates that the enhanced ensemble approach correctly identifies large majority of cervical cancer cases showing its reliability in prediction, assisting medical experts on early diagnosis. However, this study is limited to secondary datasets which involves collection of only secondary source datasets from the UCI repository for training and validating Machine Learning models. Future work should focus on advancing to Deep Learning to overcome limitation of utilising a feature extraction in Machine Learning algorithms technique and adding a lot more medical images to the dataset.