WeBear: Stock Forecasting Application

WeBear is a full-stack stock price forecasting application that simplifies stock forecasting for everyone from casual traders to data enthusiasts.

Comment

Inspiration

Our passion for finance and trading in the stock market has always pushed us to explore the intersection of technology and real-world investing. We’ve often used stock analysis tools and visualized trends manually, but we wanted to build something smarter; something that not only automates analysis but also predicts what’s ahead. That’s where we got the idea for WeBear: a sleek, full-stack application that empowers users to forecast stock prices with powerful visual insights, backed by data science and clean design.

What it does

WeBear is a stock forecasting web application that allows users to:

  • Enter a stock ticker symbol and select a date range
  • View predicted stock prices using the ARIMA model
  • Analyze moving averages and price trends
  • See confidence intervals in forecast charts
  • Interact with beautifully styled charts — zoomable on click
  • View key stats like previous high, open, close, and volume in intuitive cards

How we built it

We built WeBear using a modular full-stack architecture:

  1. Frontend: Built with React (Next.js) and TypeScript, styled using CSS Modules, and powered by Axios for API calls. The UI features a dynamic gradient background, interactive plots, custom cards, toast notifications, and date input controls.
  2. Backend: Developed using FastAPI and Uvicorn, with support for CORS. It fetches stock data via yfinance, processes it with pandas, and trains a pmdarima ARIMA model with confidence intervals.
  3. Visualization: Matplotlib is used to generate four base64-encoded charts — forecast, moving average, trend, and trend+forecast overlay.
  4. Deployment: Dockerized both frontend and backend services and used Docker Compose to bring them up seamlessly.

Challenges we ran into

  1. Choosing the Right Model: Experimenting with ARIMA and its parameters took time. I tuned p, q, d, seasonal components, and even added exogenous variables like volume for accuracy.
  2. Backend-Frontend Integration: Passing base64-encoded charts and structuring the response JSON in a clean, scalable way took careful planning.
  3. Debugging Minor Issues: Whether it was a FastAPI import path or a broken chart display due to CSS or image overflow, many small things added up.
  4. Design Decisions: Making the interface both intuitive and visually appealing required multiple iterations; from hoverable cards to animated spinners and responsive layouts.
  5. Docker Compatibility: Handling versioning issues (like with python:3.13 image), TLS errors, and ensuring production-like containers was a learning curve.

Accomplishments that we're proud of

  • Fully integrated backend + frontend, built from scratch
  • Real-time forecasts with confidence intervals and visual insights
  • Beautiful and responsive user interface with chart interactivity
  • Modular, scalable backend that can easily be extended with other models
  • Dockerized, ready-to-deploy architecture

What we learned

  • How to build and structure a product-level application from end to end
  • ARIMA modeling using pmdarima, including seasonal adjustments, exogenous variables, and forecast evaluation metrics like MAPE
  • Serving visual plots over APIs using base64
  • Frontend best practices with React/Next.js, including CSS modules and error handling
  • Deploying full-stack apps with Docker + Docker Compose
  • Most importantly, how to combine technical skills with personal interests to build something truly meaningful

What's next for WeBear: Stock Forecasting Application

WeBear is just getting started. Here’s what’s next:

  1. Model Caching: Save trained models for reuse, improving performance
  2. News Sentiment Analysis: Use NLP to integrate sentiment-driven forecasting
  3. Responsive Redesign: Optimize the UI for mobile users
  4. More Indicators: Add RSI, MACD, Bollinger Bands, etc.
  5. Live Data Support: Integrate WebSocket-based streaming for real-time updates
  6. Cloud Deployment: Deploying the whole system onto AWS for scalability
Share this project:

Updates

Submission history