This project focuses on detecting fraudulent transactions using a hybrid deep learning approach. It combines a Multi-Layer Perceptron (MLP) and an Autoencoder to improve fraud detection performance.
- Source: Kaggle - Fraud Detection Dataset
- The dataset contains credit card transactions, labeled as fraud (
is_fraud = 1) or genuine (is_fraud = 0).
📂 fraud-detection │── 📜 fraud_detection.ipynb # Jupyter Notebook with the complete workflow │── 📜 README.md # Project documentation │── 📜 requirements.txt # List of required dependencies
Clone the repository and install dependencies:
git clone https://github.com/DYNAMO_PENTESTER/fraud-detection.git cd fraud-detection pip install -r requirements.txt Steps in the Notebook Data Preprocessing
Load dataset and remove unnecessary columns
Encode categorical variables
Normalize numerical features
Handle class imbalance using SMOTE
Model Training
MLP Model: A deep learning model trained with fraud labels
Autoencoder: An unsupervised model trained on normal transactions to detect anomalies
Hybrid Model for Fraud Detection
Combines predictions from MLP and Autoencoder for better fraud detection
Evaluation & Visualization
Performance metrics: Accuracy, Confusion Matrix, ROC AUC Score
Feature importance analysis using SHAP
Precision-Recall curve for model evaluation
Results The hybrid model improves fraud detection accuracy by combining two approaches.
Precision-Recall curve analysis shows improved performance over a standalone MLP.
How to Run Execute the Jupyter Notebook step by step:
jupyter notebook fraud_detection.ipynb
Future Improvements:
Fine-tuning hyperparameters for better results
Exploring other anomaly detection techniques
Enhancing feature engineering for better fraud detection
🚀 Author Dynamo