Table 11 Metrics comparison.
From: RABEM: risk-adaptive Bayesian ensemble model for fraud detection
Reference | Method | Technique used | Accuracy | Precision | Recall | F1-score |
|---|---|---|---|---|---|---|
Blockchain and smart contract-based approach with a collaborative machine learning model | Incremental machine learning updates using blockchain-secured data with incentives | 96.83% | 95% | 96% | 96% | |
Comparison of six machine learning algorithms for fraud detection | Gradient boosting algorithm applied across multiple datasets and testing scenarios | 96.9% | – | – | – | |
Two-layer stacking prediction model combining logistic regression, SVM, and random forest | Up-sampling, under-sampling, fusion methods, and GridSearchCV for parameter tuning | 87% | – | 97% | – | |
Bayesian-optimized extremely randomized trees (BERT) for credit card fraud detection | Tree-structured Parzen Estimator (TPE) for hyperparameter tuning and optimization | – | 97% | 92% | 95% | |
Random Forest (RF) algorithm for credit card fraud detection | SMOTE for addressing an imbalanced dataset and enhancing model performance | 98% | 95% | 79% | 86% | |
Random Forest Algorithm used for credit card fraud detection | The hybrid resampling method is applied to balance the dataset for fraud detection | 97.66% | 98.85% | 95.54% | 97.37% | |
Proposed work | Bayesian ensemble model | RABEM: Bayesian ensemble model | 99.38% | 97.24% | 97.85% | 99.41% |
