用智能优化算法选特征，提升糖尿病早期诊断准确率
DiabML: AI-assisted diabetes diagnosis method with meta-heuristic-based feature selection
- 结合鲸鱼优化算法与机器学习进行特征筛选
- 采用AdaBoost分类器达86.1%准确率，优于现有方法
- 适合医疗AI研究者和糖尿病筛查系统开发者
糖尿病是一种由高血糖引发的慢性疾病，可导致肾衰竭、心梗、失明和中风等多种并发症。借助AIoMT（人工智能物联网）技术，融合物联网与机器学习，可有效实现糖尿病风险的早期检测。本文提出一种混合诊断方法DiabML，利用鲸鱼优化算法（BWO）进行特征选择，并通过SMOTE处理数据不平衡问题。实验结果表明，DiabML在使用AdaBoost分类器时达到86.1%的分类准确率，显著优于现有方法，验证了其在糖尿病早期预警中的有效性。
原文摘要 · Abstract (English)
Diabetes is a chronic disorder identified by the high sugar level in the blood that can cause various different disorders such as kidney failure, heart attack, sightlessness, and stroke. Developments in the healthcare domain by facilitating the early detection of diabetes risk can help not only caregivers but also patients. AIoMT is a recent technology that integrates IoT and machine learning methods to give services for medical purposes, which is a powerful technology for the early detection of diabetes. In this paper, we take advantage of AIoMT and propose a hybrid diabetes risk detection method, DiabML, which uses the BWO algorithm and ML methods. BWO is utilized for feature selection and SMOTE for imbalance handling in the pre-processing procedure. The simulation results prove the superiority of the proposed DiabML method compared to the existing works. DiabML achieves 86.1\% classification accuracy by AdaBoost classifier outperforms the relevant existing methods.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。