Autobot

What it does

We built 3 machine learning models for Collision, Comprehensive, and DCPD incurred losses based on client information. Some of the factors taken into account include vehicle year, driving experience, and geographic location. We also built a user-friendly Windows form interface to interact with the models.

How I built it

The models were built using Python, the Keras API, and the Tensorflow library. The UI was built in Visual Studio using Visual Basic.

Challenges I ran into

I was unable to get Tensorflow to work on my personal laptop, so I was forced to build the model on a virtual machine, which took longer. Additionally, we had trouble determining which factors were accurate predictors of incurred loss.

Accomplishments that I'm proud of

I am proud of having built a functional machine learning model that can predict DCPD incurred loss to within $2000.

What I learned

How to use Tensorflow and Keras, as well as the auto insurance industry in Ontario.

What's next for Autobot

We hope to improve accessibility by moving from a Windows application to a web application, which will be written using React. Additionally, we will integrate chatbot capabilities using Microsoft Bot Framework and LUIS.

Built With

• python
• tensorflow
• visual-basic
• visual-studio