Inspiration
In the Fintech industry, there are currently many services that offer financial institutions an analysis of whether or not a client is at-risk of defaulting on a potential loan. Many of these programs that predict loan defaulting are only available to the institutions, leaving customers in the dark. Instead, we wanted to develop a program that would predict loan defaulting, and if a loan was not approved, provide the user with potential causes and next steps to improve their chances of getting a loan next time.
What it does
The LoanSimple program uses the random forest algorithm to predict whether or not a user will be approved for a loan. If the user is not approved for a loan, the LoanSimple will give the user a couple reasons why (depending on their inputs) and some steps on how to improve their application the next time around so they can secure a loan.
How we built it
We used HTML and CSS to develop the front end of the LoanSimple program. For the backend, Python was used to create the trained machine learning model. To interface the frontend with the backend, python flask was used to deploy the model.
Challenges we ran into
The initial dataset we used to train our machine learning model was very imbalanced. As a result, our initial model was overfitting. Instead, we had to find a more balanced dataset to train our model.
Accomplishments that we're proud of
- Our machine learning model has an accuracy of 87%
What we learned
- Learned how to set up, train and test a basic machine learning model
- Learned about how to develop web apps using Flask
- Learned more about the financial industry and what factors are considered when analyzing the risk of a loan
What's next for LoanSimple
- Extend LoanSimple and train the model with data that is more closely related to the Canadian currency.
Log in or sign up for Devpost to join the conversation.