Scholarlink

Credit card fraud detection is crucial for preventing financial losses and maintaining trust. However, challenges such as class imbalance and irrelevant features hinder model performance. This study proposes a framework combining SMOTE for class balancing and Particle Swarm Optimization (PSO) for feature selection to enhance fraud detection. Evaluations using CatBoost and XGBoost achieved top AUC scores of 97.19 and 97.03%, demonstrating strong fraud detection capabilities while reducing computational costs. The PSO-based feature selection significantly improved accuracy and efficiency, making it ideal for real-world applications. This research highlights how advanced sampling and feature optimization techniques enhance fraud detection models.

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

Leveraging Swarm Intelligence to Optimize Feature Selection in Credit Card Fraud Detection

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Hayat Sahlaoui,
  • Meryem Saket,
  • Abdelati Oualla,
  • Amine Sallah,
  • Abdelaaziz Hessane,
  • Mostafa Merras

摘要

Credit card fraud detection is crucial for preventing financial losses and maintaining trust. However, challenges such as class imbalance and irrelevant features hinder model performance. This study proposes a framework combining SMOTE for class balancing and Particle Swarm Optimization (PSO) for feature selection to enhance fraud detection. Evaluations using CatBoost and XGBoost achieved top AUC scores of 97.19 and 97.03%, demonstrating strong fraud detection capabilities while reducing computational costs. The PSO-based feature selection significantly improved accuracy and efficiency, making it ideal for real-world applications. This research highlights how advanced sampling and feature optimization techniques enhance fraud detection models.