Scholarlink

Imbalanced data is crucial for contemporary machine learning models. Evaluating the effectiveness of machine learning classifiers under an unfair dataset is problematic. Several resampling techniques were applied to the unwarranted dataset and pre-processing stages to increase the accuracy of prognosis classifiers. This proposed work discusses the basics of resampling techniques such as Random under-sampling, Random oversampling, SMOTE (Synthetic Minority Over-sampling Technique), SMOTE-ENN (Synthetic minority over-sampling technique and Edited nearest neighbor), ADASYN (Adaptive Synthetic), and SMOTE-Tomek to handle this Weakness with unbalanced data. The proposed system uses an unstable credit card fraudulent dataset collected from Kaggle, which contains 21 columns and 28382 rows, of which 18497 negative samples and 4208 positive samples. The balanced dataset got by applying resampling methods SMOTE, SMOTE-ENN, ADASYN, and SMOTE-Tomek. Afterward, it uses machine learning techniques such as Logistic Regression (LR), Decision tree (DT), and Random Forest (RF) for classification. From this analysis, it selects the optimal scheme and classifier. The grey wolf optimization technique improves the classifier’s performance, which optimizes the samples after sampling. This hybrid method’s effectiveness was evaluated with standard evaluation parameters. The hybrid Random Forest technique achieved 92% accuracy from this analysis compared to hybrid 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

Handling Imbalanced Data for Credit Card Fraudulent Detection: A Machine Learning Approach

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • E. Sujatha,
  • V. Umarani,
  • K. S. Rekha,
  • P. V. Gopirajan,
  • V. Manickavasagan

摘要

Imbalanced data is crucial for contemporary machine learning models. Evaluating the effectiveness of machine learning classifiers under an unfair dataset is problematic. Several resampling techniques were applied to the unwarranted dataset and pre-processing stages to increase the accuracy of prognosis classifiers. This proposed work discusses the basics of resampling techniques such as Random under-sampling, Random oversampling, SMOTE (Synthetic Minority Over-sampling Technique), SMOTE-ENN (Synthetic minority over-sampling technique and Edited nearest neighbor), ADASYN (Adaptive Synthetic), and SMOTE-Tomek to handle this Weakness with unbalanced data. The proposed system uses an unstable credit card fraudulent dataset collected from Kaggle, which contains 21 columns and 28382 rows, of which 18497 negative samples and 4208 positive samples. The balanced dataset got by applying resampling methods SMOTE, SMOTE-ENN, ADASYN, and SMOTE-Tomek. Afterward, it uses machine learning techniques such as Logistic Regression (LR), Decision tree (DT), and Random Forest (RF) for classification. From this analysis, it selects the optimal scheme and classifier. The grey wolf optimization technique improves the classifier’s performance, which optimizes the samples after sampling. This hybrid method’s effectiveness was evaluated with standard evaluation parameters. The hybrid Random Forest technique achieved 92% accuracy from this analysis compared to hybrid models.