기계학습 기법을 사용한 신장암 병기 분류 성능의 개선

초록

Utilizing gene expression data from kidney cancer patients, we have developed a machine learning-based deep learning algorithm to extract significant genes for predicting the patients' prognosis and enhance classification performance while addressing data imbalance issues. Particularly, classification based on tumor stage plays a crucial role in determining appropriate treatment approaches for kidney cancer patients and predicting post-treatment prognosis. We classified kidney cancer tumor stages into four categories and evaluated their performance. The results demonstrated that the SVM algorithm, utilizing an autoencoder for feature extraction and addressing data imbalance through the SMOTE technique, exhibited the best performance in terms of accuracy, recall, precision, F1-score, and AUC. These results can be utilized to choose the most suitable treatment strategy at the current state and for predicting the prognosis and enabling early diagnosis of kidney cancer. © The Korean Institute of Electrical Engineers.

키워드