Multivariate Stock Forecast Dashboard (Close + Volume)

A Streamlit dashboard for multivariate stock forecasting using Close and Volume.
Supports three model options: XGBoost, Prophet, and SARIMAX. The project includes robust preprocessing (handling MultiIndex columns from yfinance, datetime normalization, dtype sanitization), lag & rolling feature engineering, recursive multi-step forecasting for tree models, and defensive code to handle common runtime issues.

Contents

• app.py / r1_ui_forecast.py — main Streamlit script (replace with your script name)
• utility modules or functions:
  • feature engineering (lags, rolling stats)
  • model wrappers for XGBoost, Prophet, SARIMAX
• requirements.txt — Python dependencies

Features

• Streamlit UI to choose ticker, date range, model, and forecast horizon.
• Multivariate modeling:
  • XGBoost: lag & rolling features, recursive multi-step forecasting, GridSearch with safeguards on small datasets.
  • Prophet: Volume used as regressor (if provided).
  • SARIMAX: Volume used as exogenous variable.
• Robust input handling:
  • Auto-detect & normalize Date, Close, Volume columns (handles MultiIndex / tuple column names).
  • Convert datetime to numeric (ordinal/dayofweek) or engineered features.
  • Sanitize predictions and arrays to ensure 1-D shapes for metrics.
• Defensive GridSearch usage — reduces/avoids heavy tuning on small datasets.
• Visualizations: holdout vs predicted, historical + forecast combined chart.

---

Quick start

1. Clone repository

git clone https://github.com/<your-username>/multivariate-stock-forecast.git
cd multivariate-stock-forecast

2. Create & activate virtual environment (recommended) macOS / Linux:

python -m venv .venv
source .venv/bin/activate

Windows (PowerShell):

python -m venv .venv
.venv\Scripts\Activate.ps1

3. Install dependencies

pip install -r requirements.txt

If you experience issues installing Prophet on Windows, try using conda:

conda install -c conda-forge prophet

4. Run the app

streamlit run r1_ui_forecast.py