---
license: mit
title: spectral analysis
sdk: gradio
emoji: 🏆
colorTo: red
---
# Enhanced Spectral Stock Analysis (Adaptive Quant Engine)

Professional-grade Hugging Face Space integrating advanced DSP-based spectral cycle analysis with a rolling adaptive backtest and Bayesian optimization of meta-parameters.

## Key Features
- Adaptive Anti-Aliasing Filter (Butterworth) with volatility and regime sensitivity
- Multi-Frequency Momentum Extraction (phase + amplitude weighted)
- Market Regime Detection (ADX-based)
- Dynamic Parameter Heuristics + Bayesian Meta-Optimization (cutoff, order, thresholds, frequency weighting)
- Rolling Window Backtest (Train/Validation) with sliding window + weekly step
- Statistical Signal Validation (forward return success rate, Sharpe, Sortino)
- Performance Metrics: Sharpe, Sortino, Hit Rate, Max Drawdown
- Live Adaptive Signal Projection using last optimized window
- Data Caching Layer for efficiency

## Project Structure
```
src/
  core/
    spectral_analyzer.py
    rolling_engine.py
    bayes_optimizer.py
    config.py
  utils/
    metrics.py
app.py
```

## Running Locally
Install dependencies:
```bash
pip install -r requirements.txt
```
Run app:
```bash
python app.py
```

## Hugging Face Space
Space type: Gradio. Set `app.py` as entry point.

## Disclaimer
Educational & research tool. Not investment advice. Use at your own risk.

## License
MIT