Performance Comparison of kNN, Random Forest and SVM in the Prediction of Cervical Cancer from Behavioral Risk
Ali Değırmencı · AYBU AVESIS · 2022
Cervical cancer is one of the most common vital diseases that still seriously affects women worldwide. Early detection of it may not be possible due to late onset of symptoms, community norms, unavailable healthcare facilities, and medical cost. Computer aided diagnostic tools have shown very successful results in the early diagnosis of diseases in recent years. Especially the developments in computer technology have increased the success of machine learning-based methods. This study presents and analyzes 3 different machine learning based algorithms (k nearest neighbor, support vector machines (SVM), and random forest) to predict cervical cancer. Hyperparameter optimization of algorithms is performed by exhaustive grid search and k-fold cross validation is used to increase the reliability of the results. Among the benchmarked methods, the best performance was obtained with the SVM method with the sigmoid kernel, and the accuracy, precision, recall, and F1-score metrics were 0.9274, 0.9093, 0.8410, 0.8565, respectively