README.md

metadata

title: Stock Price Forecast
emoji: 📈
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 5.49.1
app_file: app.py
pinned: false

Stock Price Forecasting with LSTM + MC Dropout

Predict next-close stock prices for multiple horizons (Daily, Weekly, 4-Weeks, 6-Months, 1-Year) with confidence intervals using per-ticker LSTM models and Monte Carlo dropout.

---

Using the Space

You can run the app directly in this Hugging Face Space:

1. Select a ticker (TSLA, NVDA, SPY).
2. Select a forecast horizon (1d, 1w, 4w, 6m, 1y).
3. Click Predict to see JSON output with predicted price, confidence interval, and plus-minus percent.

---

Running Locally

If you want to run the models locally, check out the included notebook:

• Price_Prediction.ipynb — demonstrates downloading models/scalers, preparing input data, and running predictions for all horizons.
• You can try your own tickers by editing the notebook and adding the tickers you want to forecast.
  Note: You will need your own Polygon API key to fetch historical stock data.
• You can also customize the HORIZON_CONFIGS dictionary to adjust model parameters for each horizon or even add new forecast horizons.

Clone the repo and install dependencies with the following:

git clone https://huggingface.co/spaces/jtassos2025/Price-Prediction
cd Price-Prediction
pip install -r requirements.txt