Comparison of Machine Learning and Deep Learning models for Cervical Cancer Prediction

Traditional cervical cancer type categorization relies heavily on the pathologist’s experience, which is inefficient and inaccurate. Colposcopy is an important part of preventing cervical cancer. Over the last 50 years, colposcopy, in combination with precancer screening and therapy, has played a critical role in lowering the incidence and death of cervical cancer. It justifies the importance of predicting cervical cancer at an early stage. For prediction, the proposed approach makes use of the UCI data repository and machine learning classifiers. The data is preprocessed, and the repository is updated with feature extraction and validations. After that, the pre-processed data is then applied to various machine learning classifiers namely Logistic Regression, Random Forest, Gradient Boosting, Support vector machine (SVM), and all its ensemble classifier. The ensemble classifier has not surpassed the results of individual classifier and thus preferred deep learning model using artificial neural network. The comparison is done in terms of accuracy and fmeasure along with recall and precision. The results revealed that the neural network model is the best with 97.67% of accuracy and 98.78% of F-score.