Scholarlink

<p>Amid the challenges of accurate classification under class imbalance, this study introduces EMGNB: a novel enhancement of Gaussian Naïve Bayes. EMGNB incorporates two key mathematical innovations; first: a dynamic enhancement factor α within the Gaussian PDF to capture sharp deviations and multi-modal distributions and second: a modified covariance matrix using partial correlation to relax naive feature-independence. Additionally, a hybrid SMOTE-ENN strategy addresses imbalance at the data level. Evaluated on a large imbalanced dataset (19k records), EMGNB achieved 99.7% accuracy, AUC &gt; 0.999, and significantly lower false negatives. Unlike previous methods that alter data or representational layers, EMGNB introduces internal probabilistic and structural improvements. These contributions position the model as both efficient and robust also ready for critical applications in fraud detection, cybersecurity, and healthcare.</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

EMGNB: an enhanced and modified Gaussian Naive Bayes algorithm for accurate classification of imbalanced big data

  • 期刊论文

  • 发表日期：

    2025年10月9日

查看全文
  • Hussam Mezher Merdas

摘要

Amid the challenges of accurate classification under class imbalance, this study introduces EMGNB: a novel enhancement of Gaussian Naïve Bayes. EMGNB incorporates two key mathematical innovations; first: a dynamic enhancement factor α within the Gaussian PDF to capture sharp deviations and multi-modal distributions and second: a modified covariance matrix using partial correlation to relax naive feature-independence. Additionally, a hybrid SMOTE-ENN strategy addresses imbalance at the data level. Evaluated on a large imbalanced dataset (19k records), EMGNB achieved 99.7% accuracy, AUC > 0.999, and significantly lower false negatives. Unlike previous methods that alter data or representational layers, EMGNB introduces internal probabilistic and structural improvements. These contributions position the model as both efficient and robust also ready for critical applications in fraud detection, cybersecurity, and healthcare.