首页|BESS: Balanced evolutionary semi-stacking for disease detection using partially labeled imbalanced data
BESS: Balanced evolutionary semi-stacking for disease detection using partially labeled imbalanced data
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NSTL
Elsevier
Machine learning offers automatic and objective approaches for disease detection based on biomedical data. However, 1) the percentage of patients in the real world is smaller than that of healthy people; 2) the annotations by medical experts are costly. Therefore, medical datasets are often imbalanced and partially labeled. To address this problem, we propose balanced evolutionary semi-stacking (BESS) for disease detection using partially labeled imbalanced (PLI) data. BESS aims to detect illnesses by considering the input of the color, texture, and geometry features extracted from tongue images. Specifically, the proposed method first mitigates the class imbalance problem and leverages the unlabeled data through the so-called balanced evolutionary co-training approach. Then BESS exploits both the data and classifier diversity obtained by balanced evolutionary co-training to improve the performance of the stacking ensemble. We quantitatively evaluate the proposed algorithm based on the PLI tongue image database. BESS achieves the best performance in detecting diabetes mellitus, chronic kidney disease, breast cancer, and chronic gastritis, compared to other state-of-the-art methods. The results of the experiments substantiate the superiority and effectiveness of the proposed algorithm. Codes and datasets have been made publicly available at url: https://github.com/CUHKSZ-NING/Balanced-Evolutionary-Semi-Stacking. (C) 2022 Elsevier Inc. All rights reserved.
NSTL
Elsevier
