Scholarlink

<p>This research study addresses the pressing issue of diabetes prediction using advanced machine learning techniques, presenting the development of an ensemble model that significantly outperforms existing methods by 1.6% using area under the curve as the primary performance metric. The proposed model achieved an area under the curve (AUC) of 0.946, surpassing the previously best-performing model, extreme gradient boosting (XB) by <b>0.7%</b>. The improved ensemble model has outrightly outperformed other existing machine learning (ML) models from the literature by <b>1.6%.</b> This improvement is particularly notable given the challenges posed by outliers and missing values in the dataset, which complicate diabetes prediction. The choice of choosing algorithms such as k-nearest neighbor (KNN), random forest (RF), AdaBoost (AB), and extreme gradient boost (XB) was meticulously informed by their distinct strengths and limitations, which are critical for the efficacy of this research study. KNN is particularly user-friendly, making it accessible for preliminary analyses; however, it exhibits sensitivity to feature scaling; RF, while robust and capable of handling a variety of data distribution, has a propensity for overfitting, which is mitigated through tuning. AB is effective in enhancing the performance of weak learners, yet it may encounter challenges with imbalanced datasets, an aspect that was addressed through soft weighted sampling vote. XB demonstrates remarkable predictive performance, bolstered by its built-in cross-validation and parallel processing; nevertheless, it remains sensitive to outliers, highlighting the importance of thorough data processing. By integrating these algorithms into an ensemble framework, this study effectively mitigated their individual limitations, leading to a more accurate and improved reliable prediction model. The innovative approach of hyperparameter tuning through randomized search further enhanced the model’s performance, marking a significant advancement in the use of artificial intelligence (AI) for diabetes prediction. </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

An improved performance model for artificial intelligence-based diabetes prediction

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年6月23日

查看全文
  • Ugwu Hillary Okwudili,
  • Oparaku Ogbonna Ukachukwu,
  • V. C. Chijindu,
  • Michael Okechukwu Ezea,
  • Buhari Ishaq

摘要

This research study addresses the pressing issue of diabetes prediction using advanced machine learning techniques, presenting the development of an ensemble model that significantly outperforms existing methods by 1.6% using area under the curve as the primary performance metric. The proposed model achieved an area under the curve (AUC) of 0.946, surpassing the previously best-performing model, extreme gradient boosting (XB) by 0.7%. The improved ensemble model has outrightly outperformed other existing machine learning (ML) models from the literature by 1.6%. This improvement is particularly notable given the challenges posed by outliers and missing values in the dataset, which complicate diabetes prediction. The choice of choosing algorithms such as k-nearest neighbor (KNN), random forest (RF), AdaBoost (AB), and extreme gradient boost (XB) was meticulously informed by their distinct strengths and limitations, which are critical for the efficacy of this research study. KNN is particularly user-friendly, making it accessible for preliminary analyses; however, it exhibits sensitivity to feature scaling; RF, while robust and capable of handling a variety of data distribution, has a propensity for overfitting, which is mitigated through tuning. AB is effective in enhancing the performance of weak learners, yet it may encounter challenges with imbalanced datasets, an aspect that was addressed through soft weighted sampling vote. XB demonstrates remarkable predictive performance, bolstered by its built-in cross-validation and parallel processing; nevertheless, it remains sensitive to outliers, highlighting the importance of thorough data processing. By integrating these algorithms into an ensemble framework, this study effectively mitigated their individual limitations, leading to a more accurate and improved reliable prediction model. The innovative approach of hyperparameter tuning through randomized search further enhanced the model’s performance, marking a significant advancement in the use of artificial intelligence (AI) for diabetes prediction.