Efficient Feature Selection Approach for Prediction of Cervical Cancer Using Machine Learning Algorithms

Diagnosing cervical cancer early is crucial for successful treatment. Early detection proves to be challenging in areas where medical infrastructure is scarce, as is the awareness and information regarding cervical cancer. Machine learning and deep learning are efficient ways of catering to the early detection of cervical cancer while providing accurate results. This study compares various machine-learning approaches for predicting cervical cancer in patients. Each model undergoes tests across various conditions, including data variations, balance adjustments, and feature selection metrics. The dataset used in this study is processed and cleaned by applying machine-learning-based imputations and data balancing techniques SMOTE and ADASYN. Feature selection metrics like ANOVA, MRMR and PCA are used to select the most relevant features for the study. Accuracy, Error Rate, Precision, Recall and Specificity are used to measure the performance of all models in this study. An enhanced KNN model along with a refined Ensemble model attained the highest performance, securing an approximate accuracy of 99.54% and an F1 Score of around 99%.