Machine Learning Based Security System

Credit Card Fraud Detection Using AI

FraudShield AI detects suspicious credit card transactions using machine learning, SMOTE balancing, XGBoost classification, and advanced model evaluation.

Try Prediction View Results
Credit card security

Best Model: XGBoost

High fraud recall, excellent ROC-AUC, and very low overfitting gap.

99.91%

Accuracy

95%

Recall

0.0009

Overfit Gap
Project Pipeline

How FraudShield AI Works

Complete machine learning workflow from dataset to fraud prediction.

Dataset

Credit card transaction records with fraud and legitimate classes.

Preprocessing

Missing values, duplicate removal, scaling, and train-test split.

SMOTE

Class imbalance handled using Synthetic Minority Over-sampling.

XGBoost

Best model selected using recall, F1 score, ROC-AUC, and overfitting gap.

Model Highlights

The system compares multiple ML algorithms and selects XGBoost as the best-performing model.

9

Models Tested

0.8636

F1 Score

0.9998

ROC-AUC