Classification of patients with stage III invasive ductal carcinoma using machine learning methods

ABSTRACT Aim: In the thesis study, it was aimed to find the method with high classification success among the methods used in the study by comparing the machine learning methods according to the classification performance. Material and Method: In our study, the data set of 302 patients with invasive ductal carcinoma, one of the breast cancer types, and 24 different data sets obtained by simulation were used to compare the classification performances of support vector machines, random forest and artificial neural network methods. The success of classifications of the methods used were compared according to the general accuracy, sensitivity, specificity, F-measure, Matthews correlation coefficient, AUC and discriminant power in breast cancer data. In the simulation data, the difference between train-test accuracy and the significance of this difference were evaluated. Results: The highest survival classifying accuracy (80%) and high performance evaluation criterion values of the test set of invasive ductal carcinoma stage III patients were obtained from radial kernel support vector machines. In the simulation data, the high classification accuracy and the small difference between accuracies were obtained from the support vector machines in general. Conclusion: Support vector machines have higher accuracy in both the real data set and simulation data than random forest and artificial neural networks.