Predict Fraud

Feature Extraction

Transaction details are converted into 28 numerical features matching the trained model's input space.

Random Forest Model

100 decision trees vote on whether the transaction is fraudulent based on patterns learned from 284,807 real transactions.

Risk Score Output

A probability score (0–1) is returned. Above 0.7 = Fraud, 0.4–0.7 = Review, Below 0.4 = Safe.

Action & Alert

In production, high-risk scores trigger instant alerts, block the transaction, and log it for review.