Scholarlink

<p>In modern healthcare, the effective management of medical data is critical for informed decision-making and patient care. Despite technological advancements, challenges remain, particularly with missing data in medical classification. Traditional imputation methods often fall short, especially for categorical data. This study proposes a novel fuzzy logic-based imputation technique, FC-KNNI, which integrates fuzzy logic principles to handle uncertainty and imprecision in both numerical and categorical medical data. We evaluate FC-KNNI’s impact on the performance of medical classification by comparing it against the classical KNN-I imputer across three single classifiers (MLP, KNN, and SVM) and three ensemble methods (Bagged SVM, XGBoost, and RF), using seven diverse medical datasets, including breast cancer, heart disease, and hepatitis. The evaluation includes standard classification metrics (accuracy, precision, recall, F1-score, AUC), along with statistical significance testing (Scott-Knott) and ranking (Borda Count). The findings demonstrate that FC-KNNI enhances prediction accuracy across all tested classifiers, and it outperformed KNN-I in terms of median accuracy for ensemble models, achieving 0.848, 0.851, and 0.866 for RF, XGB and Bagged SVM, respectively. Additionally, the KNN classifier’s accuracy raised from 0.66 to 0.70. These enhancements show how well fuzzy logic works for imputation of medical data.</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

Enhancing medical classification with fuzzy logic-based missing data imputation technique

  • 期刊论文

  • 发表日期：

    2025年8月13日

查看全文
  • Ismail Moatadid,
  • Ali Idri,
  • Ibtissam Abnane

摘要

In modern healthcare, the effective management of medical data is critical for informed decision-making and patient care. Despite technological advancements, challenges remain, particularly with missing data in medical classification. Traditional imputation methods often fall short, especially for categorical data. This study proposes a novel fuzzy logic-based imputation technique, FC-KNNI, which integrates fuzzy logic principles to handle uncertainty and imprecision in both numerical and categorical medical data. We evaluate FC-KNNI’s impact on the performance of medical classification by comparing it against the classical KNN-I imputer across three single classifiers (MLP, KNN, and SVM) and three ensemble methods (Bagged SVM, XGBoost, and RF), using seven diverse medical datasets, including breast cancer, heart disease, and hepatitis. The evaluation includes standard classification metrics (accuracy, precision, recall, F1-score, AUC), along with statistical significance testing (Scott-Knott) and ranking (Borda Count). The findings demonstrate that FC-KNNI enhances prediction accuracy across all tested classifiers, and it outperformed KNN-I in terms of median accuracy for ensemble models, achieving 0.848, 0.851, and 0.866 for RF, XGB and Bagged SVM, respectively. Additionally, the KNN classifier’s accuracy raised from 0.66 to 0.70. These enhancements show how well fuzzy logic works for imputation of medical data.