Cervical Cancer Risk Identification with Random Forest Using Web Platform

Cervical cancer continues to be a major public health concern, requiring effective risk assessment tools for early detection and prevention. This study presents a newly developed web application designed to predict cervical cancer risk using machine learning algorithms. The model was developed using secondary data from the UCI Machine Learning repository, including patient behavior, disease history, and diagnostic results from Hospital Universitario de Caracas, Venezuela. Several machine learning models-Random Forest, XGBoost, AdaBoost, Gradient Boost, and SVM-were compared, with Random Forest emerging as the best-performing model. The Random Forest model demonstrated strong performance, achieving 98% accuracy, with 0.73 for recall, 0.89 for precision, and Fl-score of 0.80, an AUC-ROC of 0.85, and a specificity of 0.99, despite the imbalance in the positive class data. Cost-sensitive learning and oversampling methods were employed to address data imbalance. This model was then integrated into a Flask-based web framework, allowing users to assess cervical cancer risk without interacting with the underlying processes. Comprehensive black box testing confirmed the application's reliability and compliance with specified requirements. Unlike previous studies focused solely on data analysis, this research successfully deploys the predictive model as an accessible web application, enhancing usability and providing a practical tool for the general public to efficiently assess cervical cancer risk, thereby supporting early detection and prevention strategies.