kryp71cgh057/auto-encoder-model

How kryp71cgh057/auto-encoder-model works

This repository is a research notebook for unsupervised financial-fraud detection, not a reusable multi-service application. Its documented workflow uses transaction data to train an auto-encoder on legitimate transactions, flags anomalies using reconstruction error and a percentile threshold, and evaluates results against fraud labels. The README also describes baseline comparison and generated reports, plots, and model artifacts. The notebook is present in the tree but its source was not excerpted, so component-level implementation details and relationships rely on README documentation.

Main entry points: auto_encoder_fraud_detection.ipynb Metric evaluation · comprehensive_report_20260320_165611.txt Comprehensive report

Components 15

Data preparation

Feature preparation auto_encoder_fraud_detection.ipynb
Data partitioning auto_encoder_fraud_detection.ipynb

Anomaly detection

Legitimate-only training auto_encoder_fraud_detection.ipynb
Auto-encoder auto_encoder_fraud_detection.ipynb
Reconstruction error auto_encoder_fraud_detection.ipynb
Anomaly threshold auto_encoder_fraud_detection.ipynb
Risk tiering auto_encoder_fraud_detection.ipynb

Evaluation and outputs

Metric evaluation auto_encoder_fraud_detection.ipynb
Baseline comparison auto_encoder_fraud_detection.ipynb
Diagnostic visualizations auto_encoder_fraud_detection.ipynb
Fraud risk report auto_encoder_fraud_detection.ipynb
Saved model and config auto_encoder_fraud_detection.ipynb
Comprehensive report comprehensive_report_20260320_165611.txt

Other components

Researcher
PaySim dataset

Connections 15

Researcher → PaySim dataset: supplies data README.md
PaySim dataset → Feature preparation: provides transactions README.md
Feature preparation → Data partitioning: prepares features README.md
Data partitioning → Legitimate-only training: selects legitimate rows README.md
Legitimate-only training → Auto-encoder: trains model README.md
Data partitioning → Auto-encoder: provides evaluation data README.md
Auto-encoder → Reconstruction error: reconstructs inputs README.md
Reconstruction error → Anomaly threshold: compares errors README.md
Anomaly threshold → Risk tiering: assigns risk tier README.md
Risk tiering → Fraud risk report: produces report README.md
Risk tiering → Metric evaluation: evaluates predictions README.md
Metric evaluation → Baseline comparison: compares baseline README.md
Metric evaluation → Diagnostic visualizations: generates plots README.md
Metric evaluation → Comprehensive report: generates report README.md
Auto-encoder → Saved model and config: saves model README.md

Drawn by GitDiagram on November 11, 2026 from the repository's file tree, README and source files. Read as Markdown ·