Techtonic - Transaction Compliance Tool

Our fintech solution uses advanced analytics, machine learning, and a multilingual chatbot to streamline compliance, detect anomalies, and ensure financial oversight in today's strict regulations.

Comment

Transaction Compliance Tool: Anomaly Detection and Explainable AI

Anomaly Detection Explainability Python NodeJS

Overview

In today’s stringent regulatory environment, businesses must manage complex compliance requirements effectively. Our comprehensive transaction compliance solution combines advanced analytics, machine learning, and a multilingual chatbot interface to streamline compliance, detect anomalies, and maintain financial oversight.

Key Functionalities

1. Fault and Anomaly Detection

  • Real-Time Monitoring: Live overview of compliance health.
  • Isolation Forest Algorithm: Identifies irregular transaction patterns and isolates anomalies early.

2. User-Specific Data Input and Adaptive Learning

  • Customizable Data Input: Add individual transaction records without uploading entire files.
  • Model Re-Training: Continuously adapts to updated datasets for improved anomaly detection.

3. Suspicious Activity Analysis

  • Comprehensive Scanning: Analyzes transaction patterns and client profiles for fraud detection.
  • Risk Assessment Models: Quantifies and prioritizes risks for effective response.

4. Machine Learning-Driven Compliance

  • Automated Processes: Frees up resources by automating compliance tasks.
  • Predictive Insights: Mitigates risks proactively with advanced analytics.
  • Continuous Learning: Adapts to new data and regulatory updates.

5. Multilingual Chatbot Integration for Financial Guidance

  • Real-Time Assistance: Open Web UI integrated with LLaMA for multilingual support.
  • Analytical Insights: Data-driven responses and financial forecasts.

Detecting Faults and Anomalies in Transaction Data

Process

  1. Data Ingestion: Securely ingest financial data in real-time or add individual transactions.
  2. Anomaly Detection: Isolation Forest algorithm retrains with each input for precision.
  3. Automated Alerts: Immediate alerts for proactive compliance actions.
  4. 95% Model Accuracy: model accuracy

Solution Overview

Our transaction compliance solution streamlines financial processes, detects compliance risks early, and enhances financial oversight by leveraging:

  • Sophisticated machine learning models.
  • Customizable data input.
  • Multilingual support: multilingual support

Technologies Used

Machine Learning

  • Programming Language: Python 3.8+
  • Libraries:
    • pandas – Data manipulation.
    • scikit-learn – Machine learning algorithms (Isolation Forest).
    • shap – Explainable AI.
    • matplotlib – Visualization.
    • pickle – Model persistence.
      ### Web App
    • Programming Language: Javascript
    • Libraries & Tools:
    • Node JS / Express - Backend server setup
    • Mongo DB - Database to store user data
    • multer, csv-writer & csv-parser - To Manipulate data recieved through http requests and feed it into our ML model
    • child_process - to run the ML model directly in our NodeJS environment
    • React - Front-end
    • Lucide Icons - Icon library

Installation & How to use

  1. Clone the repository: 

    git clone https://github.com/memastermind/dotslash-repo.git
cd final-hack

  2. Run the front-end:

    cd client
npm install
npm run dev
or
npm build

  3. Start the server, along with the ML model:

    cd ..
cd server
npm install
npm start

  4. Run the chat bot locally (it will automatically connect to our app):

  5. download llama3.2:1b

  6. run docker desktop

  7. open a new terminal instance on your pc and run:

    docker run -d -p 6969:8080 --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main

  8. The app is up and running!! Go to the url mentioned in your terminal to view it. Make sure to checkout server/transactions.csv to get an idea of the required transaction fields (this is also available in the dashboard ui).

Example Outputs

1. Anomalies Detected

via form filling or direct CSV upload: running program GIF

2. Scatter Plot

A scatter plot showing anomalies in the Amount column is displayed with anomalies in red. (Development only)

Scatter Plot

Future Enhancements

  1. Additional Features:
    • Incorporate other columns like Merchant Name, Transaction Time, and Is Fraud? for enhanced analysis.
  2. Real-Time Stream Integration:
    • Enable detection in real-time transaction streams.
  3. Interactive UI:
    • Build a front-end dashboard for easier interaction and insights visualization.
  4. Advanced Explainability:
    • Add interactive SHAP visualizations for detailed feature importance analysis.
  5. Regulatory Compliance:
    • Extend to meet compliance standards such as PCI DSS and GDPR.

Contributing

Contributions are welcome! Please fork the repository, create a branch, and submit a pull request.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Built With

Share this project:

Updates