README.md

metadata

title: Alpha Predict
emoji: 🚀
colorFrom: red
colorTo: red
sdk: docker
app_port: 8501
tags:
  - financial-analysis
  - nlp
  - sentiment-analysis
  - finbert
  - streamlit
pinned: false
short_description: Market Sentiment & Volatility Prediction using FinBERT

🚀 Alpha Predict: Market Sentiment Engine

Alpha Predict is an AI-driven financial analysis tool that leverages FinBERT (Financial BERT) to quantify market sentiment from real-time and historical headlines. It correlates sentiment with S&P 500 (SPY) performance and the VIX (Fear Index) to provide a holistic view of market psychology.

🧠 Core Features

• NLP Sentiment Analysis: Uses ProsusAI/finbert to perform high-fidelity sentiment classification on thousands of market headlines.
• Hybrid Data Fetching: Integrated with Finnhub API for live market news and price action, with a robust CSV fallback mechanism for maximum uptime.
• Predictive Indicators: Analyzes "Panic Interaction" (Sentiment x Volatility) to detect market dislocations.
• Interactive Analytics: Visualizes the relationship between news sentiment trends and price movements via Streamlit.

🛠️ Technical Stack

• UI Framework: Streamlit
• Model: FinBERT (Hugging Face Transformers)
• Data Providers: Finnhub API, Yahoo Finance (via backup)
• Deployment: Docker / Hugging Face Spaces

📂 Project Structure

• app.py: Main entry point for the Streamlit dashboard.
• src/data_fetcher.py: Handles API interactions and data resilience.
• src/processor.py: Feature engineering and sentiment batch processing.
• data/: Secure storage for historical backup data to ensure 100% availability.

🚦 Getting Started

1. API Keys: Ensure your FINNHUB_API_KEY is set in the Hugging Face Space Secrets.
2. Processing: Upon launch, the app will fetch the last 45-60 days of data.
3. Inference: FinBERT runs batch inference on the latest headlines to calculate the Sent_Mean index.

---

Note: This project was developed for academic purposes to demonstrate the application of Transformer-based models in quantitative finance.