A Comparative Analysis of Advanced Ensemble Models in Cervical Cancer Prediction

Abstract

There are no symptoms in the early stages of cervical cancer, detection can only be done through regular Papanicolaou (Pap) and Human papillomavirus (HPV) tests. However, most women are not aware of these tests, if not, shy away from taking these tests, this has led to late cervical cancer diagnosis, and now, cervical cancer is one of the most common causes of cancer deaths among women. Successful cervical cancer treatment can be improved by early diagnosis, which can be achieved by analysing potential risk factors. This paper presents the performance of two advanced ensemble models; Bagging Classifier and Adaptive Boosting (AdaBoost) Classifier in predicting cervical cancer diagnosis based on documented cancer risk factors and target variables. The models were evaluated using accuracy, sensitivity and specificity metrics. Experiments done using the Cervical Cancer Risk Factors dataset found in the University of California at Irvine (UCI) repository shows that both models achieved good accuracy levels and can thus be used in early cervical cancer detection to avoid late diagnosis that has led to massive loss of lives.