Scratch-Based SOM for Credit Card Fraud Detection

This project involves implementing a Self-Organizing Map (SOM) from scratch and applying it to detect credit card fraud. The project demonstrates how a SOM can cluster data and identify fraudulent activities using an unsupervised learning approach.

---

Project Files

1. Credit_Card_Applications.csv:

   • The dataset containing information about credit card applications, including potential fraud cases.

2. som_on_Credit_Card_Fraud_Detection.ipynb:

   • A Jupyter Notebook demonstrating the application of the SOM to the credit card dataset.

3. som_scratch.py:

   • Contains the implementation of the SOM class, which defines the core functionality of the Self-Organizing Map.
   • Also includes the TestSOM class for testing the methods and functionality of the SOM class.

4. test_som_code.py:

   • A script to run all tests for the SOM class using the TestSOM class from som_scratch.py.

---

Libraries Used

• For Scratch Implementation:
  • NumPy
  • pylab
• For Application:
  • pandas

---

How to Run the Project

1. Clone the Repository:

   git clone https://github.com/arda92a/SOM-Scratch-Based-Implementation
   cd SOM-Scratch-Based-Implementation

2. Install Dependencies: Ensure you have Python installed, then install the required libraries:

   pip install numpy pandas 

3. Run the Notebook: Launch the Jupyter Notebook to see the application in action:

   jupyter notebook som_on_Credit_Card_Fraud_Detection.ipynb

4. Test the Code: To test the implementation of the SOM class, run the test script:

   python test_som_code.py

   Example output of the test script:

   ===================== Test Start =====================
   
   --- Constructor Tests ---
   [Test Results]
   
   --- Random Weights Initialization Tests ---
   [Test Results]
   
   --- Activation Distance Tests ---
   [Test Results]
   
   --- Winner Tests ---
   [Test Results]
   
   --- Win Map Tests ---
   [Test Results]
   
   --- Train Tests ---
   [Test Results]
   
   --- Update Weights Tests ---
   [Test Results]
   
   --- Distance Map Tests ---
   [Test Results]
   
   ===================== All Tests Completed =====================
   

---

Example Output

Winner Map Visualization

The winner map generated by the SOM highlights fraudulent activities. Fraudulent entries appear as white nodes, indicating anomalies in the data.

Example Map:

In the map, clusters represent different data segments, and fraud cases stand out due to their distinct characteristics.

---

Key Features

• Scratch Implementation:

  • A complete SOM built from scratch in Python (som_scratch.py), demonstrating core concepts of unsupervised learning.

• Fraud Detection:

  • Identifies fraudulent credit card applications based on clustering anomalies.

• Testing Suite:

  • The TestSOM class (in som_scratch.py) ensures that the methods in the SOM class are functioning correctly, and test_som_code.py runs all these tests.

---

Future Improvements

• Parameter Tuning:

  • Automate the selection of SOM parameters (e.g., grid size, learning rate).

• Dataset Expansion:

  • Apply the model to larger and more diverse datasets for improved robustness.

• Enhanced Visualization:

  • Add interactive visualizations for better interpretability.

---

Contributions

Feel free to contribute to this project by submitting pull requests or reporting issues. Suggestions for improvement are always welcome!

---

Contact

For any questions or feedback, please reach out to the project maintainer at ardaoztuner3@gmail.com.