Prophecy | Devpost

Inspiration

Recognizing the significant barrier to financial inclusion faced by underbanked and unbanked individuals, we sought to empower them with data-driven insights. Inspired by the potential of machine learning, we leveraged Kaggle's Credit Home Default Risk model to develop Prophecy, a LightGBM-based prediction model. This tool enables users to input their financial information and receive predictions on loan approval rates or default risk, empowering them to make informed decisions about their financial future.

What it does

Prophecy is a user-friendly tool designed to empower individuals, particularly those underbanked or unbanked, with data-driven financial insights. By analyzing user-provided financial and demographic data, Prophecy predicts the likelihood of loan approval for new applicants and assesses the risk of default for existing loan holders, providing actionable predictions without requiring extensive financial knowledge and increasing accessibility.

How we built it

To build Prophecy, we leveraged the power of LightGBM, a highly efficient gradient-boosting framework, to train our model on Kaggle's Home Credit Default Risk Dataset. This dataset, rich in relevant financial information, provided the foundation for accurate predictions.

Our tech stack, including FastAPI, React.js, and TypeScript, ensured a seamless and efficient user experience. FastAPI's lightweight and high-performance nature made it ideal for building the backend API, responsible for serving machine learning predictions. React.js and TypeScript, on the frontend, provided a dynamic and reliable user interface, allowing users to easily input data and receive timely predictions. This combination of powerful tools and technologies enabled us to create a robust and user-friendly financial prediction tool.

Accomplishments that we're proud of

Throughout Prophecy's development, we achieved significant milestones: a high-accuracy prediction model, a robust and scalable backend, and a user-centric design. By overcoming technical challenges through collaborative teamwork, we created a tool that empowers individuals with data-driven financial insights and promotes financial literacy through interactive modal viewpoints and quick resources.

What we learned

Building Prophecy was a rewarding journey that taught us the importance of clear goals, teamwork, and effective communication. We learned how to break down complex problems into smaller, manageable tasks and how to work together efficiently to achieve our objectives. By prioritizing user needs and designing a user-friendly interface, we created a tool that empowers individuals to make informed financial decisions.

What's next for Prophecy

Prophecy aims to evolve into a more dynamic and user-centric platform, expanding its capabilities to better serve underbanked and unbanked communities. By introducing new features and enhancements, the platform will provide more tailored insights, improve accessibility, and offer greater flexibility for managing financial decisions. Key upcoming developments include:

Regional and Geographic Distribution Integration: Tailor tips and insights based on users' regions to provide more relevant advice.
Personalized Educational Resources: Empower users with targeted financial education to enhance their understanding and decision-making.
Multiple Loan Functionality and User Account Storage: Enable users to manage multiple loans and store account information for a streamlined experience.

Built With

fastapi
lightgbm
pinata
postgresql
railway
react
typescript

Submitted to

HackUTD 2024: Ripple Effect

Created by

I trained and integrated LightGBM framework model for prediction with dataset, as well as integrated FastAPI backend with React frontend. It was my first time debugging and integrating various components properly, but I learned a lot about project management and the steps we should take to integrate.

Anushrutha Boyapati
I worked on the backend as well as the integration of the front-end. Creating the model was the tough part as we were trying to predict with transformed training data to fit our needs for prediction. It was my first time doing a lot of routing on my own and dealing with financial data. However, I learned my own strengths and weaknesses and what to work on moving forward :).

Victor Sim
I worked on implementing and refining the React front-end and successfully integrating it with the FastAPI backend. This involved designing dynamic user interfaces using React's component-based architecture, managing state with hooks, as well as troubleshooting routing issues with the API endpoints. It was very interesting and rewarding to work with a machine learning model for the first time!

Tejas Ramanujam
Vladislav K

Updates

Vladislav K started this project — Nov 16, 2024 11:01 PM EST

Leave feedback in the comments!

Log in or sign up for Devpost to join the conversation.