Fraud Model Trainer

Train and evaluate ensemble fraud detection models

Model Architecture

Hyperparameters

Training Data

Total Samples 2,450,000
Fraud Rate 1.2%
Features 128
Train/Val/Test 70/15/15

Training Progress

Ready

Confusion Matrix

ROC & PR Curves

Model Comparison

Model AUC-ROC PR-AUC Precision Recall F1 Latency Deploy
XGBoost Ensemble ★ 0.987 0.842 0.923 0.891 0.907 12ms
LightGBM 0.983 0.824 0.908 0.878 0.893 8ms
Neural Network 0.979 0.812 0.895 0.912 0.903 45ms
Isolation Forest 0.945 0.723 0.812 0.856 0.834 5ms