sinchumahima/fraud-detection

Credit Card Fraud Detection

This project implements an end-to-end machine learning pipeline to detect fraudulent credit card transactions using the Kaggle Credit Card Fraud Detection dataset.

Features

• Exploratory Data Analysis (EDA): Quick analysis of class imbalance and feature distributions.
• Robust Preprocessing: Scaling of Time and Amount using RobustScaler to handle outliers.
• Multiple Models: Includes Logistic Regression (baseline) and Random Forest Classifier.
• Advanced Evaluation: Focuses on Precision-Recall AUC, Confusion Matrices, and F1-scores to handle highly imbalanced data (0.17% fraud).

Project Structure

• src/: Core logic
  • data_preprocessing.py: Data cleaning and splitting.
  • model.py: Training functions for LR and RF.
  • model_evaluation.py: Performance metrics and visualization.
• main.py: Orchestrates the full pipeline.
• predict.py: Script for loading a saved model and making predictions.
• eda.py: Preliminary data exploration.
• models/: Directory containing saved .pkl model files.

How to Run

1. Install dependencies:

   pip install -r requirements.txt
   

2. Run the full pipeline:

   python main.py
   

3. Run EDA:

   python eda.py
   

4. Run the Streamlit Dashboard (Data Scientist View):

   streamlit run app.py
   

5. Run the Premium Analytics Frontend (Business/Production View):
   • First, start the API:

     python api.py
     

   • Then, open frontend/index.html in your browser.

Results

• Logistic Regression: High recall (92%) but low precision (6%).
• Random Forest: High precision (96%) and good F1-score (84%).

Artifacts

• class_distribution.png: Visualization of the imbalance.
• confusion_matrix.png: Last trained model's confusion matrix.
• precision_recall_curve.png: PR curve for performance evaluation.