We wanted to use BrainJS, and data analysis, to create a Web-app that helped banks on the local to even national level. This app allows banks to enter general information about customers, and find out, based on neural networks and machine learning, if they would pay back the loan in full. This app's inspiration mainly comes from the fact that the banking system has a huge flaw related to its customers. The fact that we have had economic turmoil due to customers abusing the bank system is reason enough to help the industry with technology that can allow them to prevent the past from happening again.
What it does
The web app starts off with three buttons; problem, evidence, and activity. The problem button gives a very brief overview of the problem with bank loans. The evidence button shows data visualizations to better illustrate the relationships between customer and loan information. With the power of Microstrategies' API and Dossier creator, we can show the relationships easily, and embed them in our web-app with ease. Finally, the activity button shows 6 inputs, and users enter the information, and the neural network will predict if you will pay off your loan in full or not.
How we built it
Challenges we ran into
Our biggest challenge was everything to work together. We had many of our micro services done. But then configuring it into our web framework was the most difficult. We got hit with a bunch of bugs and errors that we had to debug. Also learning new technologies like brain.js had its learning curve.
Accomplishments that we're proud of
We’re proud that we were able to integrate a lot of our different services together. Like using Micro Strategy's data visualization tool and integrating brain.js in the backend.
What we learned
We learned how to use brain.js and how to configure a deep learning network. We learned how to parse and clean data so it would be ready to be sent into the network. We learned how to work with node.js as our backend.
What's next for Loan Risk Mitigation with Brain JS
What’s next is to continue polishing our model. Our model’s error rate is still not where we want it. We want to build a more complex model with more usable features. We also want to continue to polish up on the design of the website.