Scholarlink

<p>In the digital era, protecting sensitive data and personal information has become increasingly difficult due to the surge in fraudulent activities across diverse sectors such as law enforcement, government, finance, and banking. This has emphasized the urgent need for advanced fraud detection methods. To address this, this paper proposes a Random Light Gradient-based CatBoost Ensemble with Enhanced Gold Rush Algorithm, tailored to accurately and efficiently identify credit card fraud activities. The proposed model leverages dynamic kernel principal component analysis to extract the most relevant features, enhancing the difference between genuine and fraudulent transactions. The detection pipeline integrated four key components: CatBoost to handle complex feature interactions, Light Gradient Boosting Machine to reduce false positives, a weighted average ensemble to evaluate predictions, and a Random Forest to manage large and intricate datasets. Performance is further optimized through the integration of the gold rush optimization algorithm, which employs a randomized strategy to reduce computational overhead. Experimental validation is conducted on multiple benchmark datasets from Kaggle, which include the financial fraud detection dataset, the credit card fraud detection dataset, and credit card.csv. The proposed approach achieved an accuracy of 99.1%, a recall of 97.3%, and a precision of 98.1%, which significantly enhances fraud detection and strengthens the security of digital transactions.</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

Optimizing Network and Systems Management for Fraud Detection: A High-Performance Approach Using Random Light Gradient-Based CatBoost Ensemble with Enhanced Gold Rush Algorithm

  • 期刊论文

  • 发表日期：

    2025年8月26日

查看全文
  • Rakan A. Alsowail

摘要

In the digital era, protecting sensitive data and personal information has become increasingly difficult due to the surge in fraudulent activities across diverse sectors such as law enforcement, government, finance, and banking. This has emphasized the urgent need for advanced fraud detection methods. To address this, this paper proposes a Random Light Gradient-based CatBoost Ensemble with Enhanced Gold Rush Algorithm, tailored to accurately and efficiently identify credit card fraud activities. The proposed model leverages dynamic kernel principal component analysis to extract the most relevant features, enhancing the difference between genuine and fraudulent transactions. The detection pipeline integrated four key components: CatBoost to handle complex feature interactions, Light Gradient Boosting Machine to reduce false positives, a weighted average ensemble to evaluate predictions, and a Random Forest to manage large and intricate datasets. Performance is further optimized through the integration of the gold rush optimization algorithm, which employs a randomized strategy to reduce computational overhead. Experimental validation is conducted on multiple benchmark datasets from Kaggle, which include the financial fraud detection dataset, the credit card fraud detection dataset, and credit card.csv. The proposed approach achieved an accuracy of 99.1%, a recall of 97.3%, and a precision of 98.1%, which significantly enhances fraud detection and strengthens the security of digital transactions.