Comparison of Performance of Machine Learning Algorithms for Cervical Cancer Classification

Abstract

Cervical cancer, which is the fourth leading cause of mortality among women, displays no symptoms in its early stages. Cervical cancer is currently diagnosed using only a few approaches using Machine Learning techniques. Certain approaches such as PAP Test, HPV Test, Colposcopy and Biopsy require medical staff intervention and cancer is not detected until a certain stage is reached. These procedures are also too costly in developing countries. Detection of Cervical Cancer using Machine Learning and Deep Learning techniques come into play to solve this issue. A few to name are: CervDetect <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[1],</sup> a hybridized model using a combination of Random Forest and Shallow Neural Networks, ResNet50 – A Convolutional Neural Network’s pre-trained model works effectively on classification of cervical cancer cells using images. This research paper experiments and analyses two Support Vector Machine (SVM) techniques as well as K-Nearest Neighbor (KNN), Random Forest(RF), Logistic Regression and Gaussian Naïve Bayes (GNB) algorithms for cervical cancer diagnosis. The dataset used is Cervical cancer (Risk Factors) Data Set from UCI Repository <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[2]</sup> . There are 32 risk factors and four target variables in cervical cancer dataset: Citology, Hinselmann, Schiller and Biopsy. The two SVM-based techniques namely SVM Linear and SVM Radial, KNN, RF, Logistic Regression and GNB have diagnosed and categorized all four targets respectively. Following that, a comparison between these six methods is done and inferences are drawn on which algorithm performs better on each of the targets.