Prediction of Renal Cell Carcinoma Based on Ensemble Learning Methods

Abstract

Objective: In recent years, ensemble learning methods have gained widespread use for early diagnosis of cancer diseases. In this study, it is aimed to establish a high-performance ensemble learning model for early diagnosis and classification of renal cell carcinomas.Methods: In the study, hemogram and laboratory data of 140 patients with renal cell carcinoma and 140 patients without renal cell carcinoma were included in the study. The data set includes 27 predictors and 1 dependent variable. The data were obtained retrospectively. In the study, classification performances of machine learning methods and ensemble learning methods were compared. In the study, classification performances of boosting, bagging, voting and stacking ensemble learning methods as well as IB1, IBk, Kstar, LWL, REPTree, Random Forest and SMO classifiers were compared.Results: REPTree classifier provided the highest performance among machine learning methods (Accuracy = 0.867). Among the ensemble learning methods, the Stacking ensemble learning method provided the highest performance in Model 6 (Accuracy = 0.906). Stacking ensemble learning methods performed higher than boosting, voting, bagging ensemble methods and machine learning methods.Conclusion: Stacking ensemble learning methods provide successful results in the early diagnosis of renal cell carcinomas. Stacking ensemble learning methods can be used as an alternative to existing methods for diagnosing renal cell carcinoma. In order to further increase the classification performance of the stacking ensemble learning method, it is recommended to choose a meta classifier suitable for the data set and variable types.