https://github.com/mohit1106/fraud-detection

an anomaly detection system on 284,807 transactions, achieving an AUC of ~0.972 with CNNs and Autoencoders.
https://github.com/mohit1106/fraud-detection

autoencoders cnn-model isolation-forest keras python scikit-learn tensorflow

Last synced: about 23 hours ago
JSON representation

an anomaly detection system on 284,807 transactions, achieving an AUC of ~0.972 with CNNs and Autoencoders.

• Host: GitHub
• URL: https://github.com/mohit1106/fraud-detection
• Owner: mohit1106
• Created: 2025-02-02T18:17:46.000Z (1 day ago)
• Default Branch: master
• Last Pushed: 2025-02-02T18:21:33.000Z (1 day ago)
• Last Synced: 2025-02-02T19:25:28.601Z (1 day ago)
• Topics: autoencoders, cnn-model, isolation-forest, keras, python, scikit-learn, tensorflow
• Language: Jupyter Notebook
• Homepage:
• Size: 954 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Fraud-Detection-in-Financial-Transactions

an anomaly detection system on 284,807 transactions, achieving an AUC of ~0.972 with CNNs and Autoencoders.

🔍 Detecting fraudulent transactions using anomaly detection models

🚀 Key Features:

Developed an anomaly detection system on 284,807 transactions, achieving an AUC of ~0.972 with CNNs and Autoencoders.

Applied oversampling, undersampling, and class-weight adjustments, improving fraud detection precision by 15%.

Optimized hyperparameters via grid search, reducing false positives by 18% and enhancing fraud detection efficiency.

🛠 Tech Stack: Python, TensorFlow, Keras, Scikit-Learn, Isolation Forest, LOF, Autoencoders, CNNs, PCA

📁 Repo Structure:

data/ → Preprocessed financial transaction dataset

models/ → Trained fraud detection models

notebooks/ → Jupyter notebooks for training and evaluation