SignalStreet

Description

SignalStreet is a financial intelligence system designed to make stock prediction more reliable, transparent, and beginner-friendly.

Instead of only outputting a BUY, SELL, or HOLD signal, our system goes one step further: it evaluates how trustworthy that prediction actually is with a confidence factor.

In real financial markets, even small fluctuations in price data can cause machine learning models to become overconfident or unstable. This leads to misleading predictions that look accurate but break under real-world volatility.

To solve this, we introduce a robustness layer on top of traditional predictive models, inspired by stress-testing techniques used in Wall Street risk management.

This layer helps users understand not just what the model predicts, but also how fragile or stable that prediction is under market uncertainty.

TechStack

Front End
React JavaScript CSS

Back End
Flask Python

Machine Learning
Numpy Pandas

What we learned

We learned how to design a full-stack application by integrating a React frontend with a Flask backend, while also navigating the challenges of turning machine learning models into usable, real-time systems. Through this process, we saw how small fluctuations in data can significantly impact model stability and prediction reliability, leading us to think beyond simple outputs and incorporate robustness and stress-testing concepts to better evaluate the trustworthiness of predictions.

TeamMates

Kazi Hossain
Aquila Nuzhat

What's next for SignalStreet

Connect the frontend and backend into a fully integrated real-time system
Improve model performance using deep learning
Add sentiment analysis from financial news
Connect to more real-time data sources such as professional market APIs

Built With

• css
• flask
• javascript
• numpy
• pandas
• python
• react