Scholarlink

<p>Although neural networks show promise in terms of accuracy, they often have trade-offs in other metrics of performance, particularly when dealing with small tabular datasets. Acknowledging the limitations of training neural networks with small datasets, earlier studies recommend the adoption of ensemble algorithms. The findings underscore the potential of ensemble models in enhancing diagnostic accuracy while maintaining balance across crucial performance metrics, offering a promising avenue for advancing disease prediction methodologies. In this study, we’ve developed a novel deep ensemble algorithm for diagnosing liver and diabetes diseases. The datasets utilized in this study are sourced from Kaggle, a widely recognized platform for datasets. The experimental results demonstrate that the proposed ensemble model performs better than most of the conventional machine-learning methods, demonstrating higher accuracy, precision, recall, and F1 scores in accurately predicting liver and diabetes diseases. Additionally, the reliability of the model is reinforced through calibration plots, probability distribution analysis, and threshold-based performance curves, which highlight its strength across different decision regions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Calibration-Boosted Self-adaptive Deep Ensemble for Disease Diagnosis

  • 期刊论文

  • 发表日期：

    2026年2月4日

查看全文
  • Ifra Altaf,
  • Sheikh Amir Fayaz,
  • Majid Zaman

摘要

Although neural networks show promise in terms of accuracy, they often have trade-offs in other metrics of performance, particularly when dealing with small tabular datasets. Acknowledging the limitations of training neural networks with small datasets, earlier studies recommend the adoption of ensemble algorithms. The findings underscore the potential of ensemble models in enhancing diagnostic accuracy while maintaining balance across crucial performance metrics, offering a promising avenue for advancing disease prediction methodologies. In this study, we’ve developed a novel deep ensemble algorithm for diagnosing liver and diabetes diseases. The datasets utilized in this study are sourced from Kaggle, a widely recognized platform for datasets. The experimental results demonstrate that the proposed ensemble model performs better than most of the conventional machine-learning methods, demonstrating higher accuracy, precision, recall, and F1 scores in accurately predicting liver and diabetes diseases. Additionally, the reliability of the model is reinforced through calibration plots, probability distribution analysis, and threshold-based performance curves, which highlight its strength across different decision regions.