Communications of the ACM
ML Algorithms Can Be Improved Significantly for Medical Purposes
The researchers built cost sensitivity into each of the algorithms. This means the algorithm gets a much bigger penalty for telling a sick person in the dataset that they are healthy, than the other way round. In medical terms, the algorithms get bigger p
Credit: Timofeev Vladimir
Ibomoiye Domor Mienye and Yanxia Sun at South Africa's University of Johannesburg have demonstrated that machine learning (ML) algorithms can be enhanced significantly for medical purposes.
The researchers used supervised binary classification algorithms with built-in cost sensitivity, meaning they received bigger penalties for providing false negatives.
Mienye and Sun used public learning binary datasets for diabetes, breast cancer, cervical cancer, and chronic kidney disease (CKD).
The Random Forest algorithm had diagnostic precision for CKD at 0.972 and recall at 0.946 out of a perfect 1.000; cost-sensitivity improved its precision at 0.990 and recall at a perfect 1.000.
Three other algorithms' CKD recall also rose from high scores to a perfect 1.000, while different programs improved differently with the other disease datasets.
From News-Medical Life Sciences
View Full Article
Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA
No entries found