An AI-powered dashboard that detects fraudulent insurance claims using Machine Learning, Deep Learning, and Anomaly Detection.
This project demonstrates how data science can help insurance companies identify suspicious claims and reduce fraud losses.
Insurance fraud costs companies billions every year. Detecting fraud manually is slow and inefficient.
This project builds an AI fraud detection system that:
- Analyzes insurance claim data
- Predicts if a claim is fraudulent
- Detects anomalies in claim behavior
- Visualizes fraud analytics in a dashboard
The application includes a Streamlit dashboard where users can input claim data or upload a dataset to detect fraud automatically.
- Python
- Pandas
- Scikit-learn
- TensorFlow / Keras
- Streamlit
- Plotly
- Joblib
This project uses multiple AI techniques:
Used for the main fraud detection model.
A neural network that learns patterns in claim data.
Detects anomalies and suspicious patterns in claims.
insurance-fraud-detection │ ├── dataset/ │ └── insurance_claims.csv │ ├── model/ │ ├── fraud_model.pkl │ ├── anomaly_model.pkl │ └── deep_fraud_model.h5 │ ├── train_model.py ├── app.py ├── requirements.txt └── README.md
✔ Fraud claim prediction
✔ Deep learning fraud scoring
✔ Anomaly detection
✔ Interactive dashboard
✔ CSV batch fraud detection
✔ Fraud analytics charts
✔ Dataset validation
✔ Download sample dataset
The model expects the following columns:
Age, Sex, VehiclePrice, AccidentArea, Fault, PastClaims
Example:
Age,Sex,VehiclePrice,AccidentArea,Fault,PastClaims 25,Male,20000,Urban,Policy Holder,1 40,Female,15000,Rural,Third Party,0 30,Male,22000,Urban,Policy Holder,2
Clone the repository:
git clone https://github.com/ASKHATDIGRASKAR/Insurance-Fraud-Detection-Using-Machine-Learning.git
Navigate to the project folder:
cd insurance-fraud-detection
Install dependencies:
pip install -r requirements.txt
Run the training script:
python train_model.py
This will generate the trained models inside the model/ folder.
Start the Streamlit application:
streamlit run app.py
Open the dashboard in your browser:
The dashboard allows users to:
- Enter claim information
- Predict fraud probability
- Upload claim datasets
- Detect high-risk claims
- Visualize fraud analytics
- SHAP explainable AI
- Fraud risk gauge visualization
- Real-time fraud detection API
- Larger fraud dataset training
- Deployment using Docker or cloud services
Contributions are welcome.
- Fork the repository
- Create a new branch
- Submit a pull request
Akshat Digraskar (0246AL231009) Akshay Parmar (0246AL231010) Alok Mishra (0246AL231011) Anand Vaidya (0246AL231012)
This project is licensed under the MIT License.