YouFraud

Real-time fraud detection system. This application is supposed to act as a validator for transactions. Transaction data would be passed in real-time and transaction validity would be determined in the context of fraudulency. Fraud transactions would be prompted to be terminated. Primary target of this application are banking systems.

Usage

• The web interface can be accessed here.

• From github:

  git clone https://github.com/MetalInMyVeins/YouFraud
  cd YouFraud/
  pip install --no-cache-dir -r app/requirements.py
  streamlit run app/main.py

• Using docker:

  docker pull metalinmyveins/you_fraud-0.1.0:v1
  docker run -p 8501:8501 you_fraud-0.1.0

• The model is also hosted here.

• The dataset can be downloaded from here.

How Does It Work

We trained a machine learning model with a transaction dataset containing 6 Million+ transaction information performing various statistical analytical methods on it. The model can predict fraud transactions with 98% accuracy. The pipeline is presented below.

Project checklist

Problem Identification

• Choosing real world problem
• Outline the project goal
• Identify primary audience

Data Collection

• Web scraping
• API integration
• Public datasets
  • Kaggle
  • UCI
  • data.gov
  • Bangladesh Bank

Exploratory Data Analysis

• Create data summary report
• Develop dashboard for data visualization

Data Preparation

• Split data into training set and test set

Apply data preprocessing techniques on training data:

• Data Cleaning

  • Handling missing values
  • Outlier detection and treatment
  • Data normalization or standardization
  • Encoding categorical variables

• Feature Engineering

  • Create new features
  • Transform existing features

• Feature Selection

  • Identify relevant features
  • Remove irrelevant features

• Document data preprocessing steps

• Build a pipeline for data preprocessing to use the same pipeline on test data

Model Development and Evaluation

• Create baseline model
• Experiment with different ML algorithms
• Evaluate models with appropriate metrics
• Perform hyperparameter tuning and model optimization
• Implement ensemble methods for improved performance
• Document model development process

Build Application

• Develop user-friendly interface
  • Use Flask, FastAPI, Streamlit, Django

Model Deployment

• Share trained model:
  • Hugging Face
  • Tensorflow Hub
• Deploy application:
  • Heroku
  • AWS
  • Azure
  • Google Cloud Platform
  • Docker
  • Kubernetes

License

See License for more information.