Analysis of Cervical Cancer using Supervised Machine Learning Classifiers and Curve Fitting

Abstract

Prognosis of cervical cancer on the basis of supervised machine learning classifiers plays a pivotal role in medical informatics. The significant factors leading to cervical cancer are on number of pregnancy, Age, Hormonal Contraceptives and Smoke.For the prediction purpose, the classifier algorithms i.e. BayesNet, SMO, Lazy-IBK, Decision Table and J48 have been applied and for each classifier the results have been marked on the basis of – (i) without preprocessing (ii) SMOTE with resampling and (iii) Class balancer. It has been observed that IBK classifier algorithm is having highest accuracy with 99.7%. The kappa statistic is 0.9935, precision is 0.997 and recall is 0.997. Based on curve fitting with respect to each parameter versus target, the Curve fitting of Hormonal Contraceptives versus Target is based on polynomial expression of degree 5.

Keywords: Cervical Cancer, BayesNet, SMO, Lazy-IBK, Decision Table, J48, curve fitting.