A New Enhanced Predictive Model for Cervical Cancer Classification Based on CatBoost and CNN Algorithms

Abstract Cervical cancer is a disease affecting women worldwide, including in impoverished countries, and it poses a significant threat to their reproductive health. The main drawback of this cancer is that it shows no early signs. Additionally, social factors and individual errors during manual testing hinder screening for this disease. Therefore, developing a model using machine learning and deep learning is a significant and trustworthy option that offers greater support than traditional detection methods. In this research, we use two powerful algorithms, Categorical Boosting (CatBoost) and Convolutional Neural Network (CNN) to tackle the problem of cervical cancer classification. Two datasets are used, Cervical Cancer Risk Factors (CCRF) and Malhari, where the first dataset is textual data. After applying the CatBoost algorithm, we reach an accuracy of 100% for CCRF dataset. The second dataset has two types of images (Pap Smear and Colposcopy), where we reach an accuracy of 87.2% for the Pap Smear type and 91.1% for the Colposcopy type by applying the newly developed CNN algorithm.