Scholarlink

<p>Credit card fraud detection has emerged as a critical issue because of the extremely skewed data of transactions and the existence of redundant or non-critical features that influence model performance. This paper presents an ideal feature selection model, which was developed around the Feature Selection using Osprey Optimization Algorithm, a nature based optimization algorithm, and a stacking ensemble classifier. The proposed method would apply a stacking ensemble of Naive Bayes, Decision Tree, Support Vector Machine and Random Forest classifiers as the base learners with the Logistic Regression as the meta-learner. The ensemble-based assessment allows identifying the relevant features strongly and efficiently can cope with class imbalance. Results of experimental work with a real-world credit card fraud dataset show that the suggested approach enhances the performance of fraud detectors in terms of precision, recall, F1-score and Matthews correlation coefficient and reduces the dimensionality of features. The outcomes prove the efficiency of the advanced algorithm on the improvement of the classification results with imbalance financial dataset.</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

A nature-inspired osprey optimization based feature selection framework with stacking ensemble model for imbalance-aware credit card fraud detection

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月9日

查看全文
  • Bharat Kumar Padhi,
  • Bighnaraj Naik,
  • Rajat Kumar Sahu,
  • Rohit Kumar,
  • Debahuti Mishra

摘要

Credit card fraud detection has emerged as a critical issue because of the extremely skewed data of transactions and the existence of redundant or non-critical features that influence model performance. This paper presents an ideal feature selection model, which was developed around the Feature Selection using Osprey Optimization Algorithm, a nature based optimization algorithm, and a stacking ensemble classifier. The proposed method would apply a stacking ensemble of Naive Bayes, Decision Tree, Support Vector Machine and Random Forest classifiers as the base learners with the Logistic Regression as the meta-learner. The ensemble-based assessment allows identifying the relevant features strongly and efficiently can cope with class imbalance. Results of experimental work with a real-world credit card fraud dataset show that the suggested approach enhances the performance of fraud detectors in terms of precision, recall, F1-score and Matthews correlation coefficient and reduces the dimensionality of features. The outcomes prove the efficiency of the advanced algorithm on the improvement of the classification results with imbalance financial dataset.