Ensemble learning methods based prediction of renal cell carcinoma

Objective: The aim of this research was to predict RCC disease by using ensemble learning methods obtained by combining different machine learning methods. It was aimed to create an ensemble learning model that provides the highest classification performance by applying different classifiers and combination of techniques to ensemble learning methods in the classification of RCC. Material and Methods: The data set was obtained retrospectively from Urology Clinic of KSU, Health Practice-Research Hospital. The sample size consisted of 280 patients. The data set includes 28 predictors and one dependent variable. In the prediction of RCC ensemble learning methods of boosting, bagging, voting and stacking were applied. In the ensemble learning methods, IB1, IBk, KStar, LWL, random forest, REPTree and SMO classifiers were used. Results: The model combining IB1, IBk, KStar, LWL, random forest and REPTree classifiers provided the highest classification performance in predicting RCC in the stacking ensemble learning method. The classification performance criteria of the model were obtained as: accuracy value 0.906, sensitivity value 0.906, precision value 0.906, specificity value 0.910 and AUC value 0.944. Conclusion: Ensemble learning methods have been successful in predicting RCC. The ensemble method that classifies RCC with the highest performance has been the Stacking ensemble learning method. The implementation of the stacking ensemble learning method in the process of knowledge discovery in medical data yields successful results. In ensemble learning methods, the development of combination of techniques and the inclusion of appropriate classifiers in the model will improve classification performance.