Inspiration
We built the app because we wanted users to be able to easily identify whether their claim is likely to be fraudulent or not, and receive reliable advice right away on what to do if it is. Being Gen Z's who are new to "adulting", we wanted to keep the process as simple and similar to a regular insurance claim report as possible to avoid getting confused by the complexity of insurance.
What it does
Our app uses a ML model to predict whether a vehicle insurance claim is likely to be fraudulent or not based on the user-provided data from the insurance claim report. If the app detects fraud, then we implemented explainable AI to explain what parameters / factors seemed unusual to the model. It then presents advice provided by generative AI on what to do based on the certain reason why the claim was thought to be fraudulent.
How we built it
We used Streamlit to build the frontend and Python and used various Python ML libraries (pandas, numpy, etc) to clean and process our dataset and also to build our Logistic Regression Model. We also used Google Gen AI APIs to provide recommendations to the user on next steps to take for a possible fraudulent claim.
Challenges we ran into
We had some trouble understanding how to create a model to predict insurance fraud, since there are many types of fraud so it can be pretty broad to think about. We had to go through lots of datasets to find one that made sense for the problem we were trying to solve.
Accomplishments that we're proud of
We were able to integrate all components of our app end-to-end.
What we learned
We learned how to use Streamlit and various generative AI APIs. We learned how to present custom-built ML model predictions in a Streamlit frontend.
What's next for Stake Your Claim
We are going to provide insurance claim fraud detection for multiple types of insurance other than just for vehicles.
Built With
- ai
- python
- streamlit
Log in or sign up for Devpost to join the conversation.