Tags: proccl/quantforecast

Release v1.1.1: Classification model with rolling predictions

Key features:
- Classification model with prob_up output
- Rolling prediction (5 historical -> 5 future days)
- Walk-forward CV optimization
- Strategy parameter optimization
- Data visualization with confusion matrix

Release v1.1.0: Walk-forward CV optimization, modular architecture, r…

…eal-time data update

Release v1.0.0 - Stable PatchTST Quantitative Trading System

Features:
- PatchTST Transformer model for stock prediction
- Feature scaling fix (price normalization + log transform)
- Bayesian hyperparameter optimization (Optuna)
- Complete backtesting with risk metrics
- Dark theme visualization
- Xiaomi (1810.HK) real data support

Performance (backtest):
- Strategy return: -15.60%
- Buy & hold: -9.03%
- Sharpe ratio: -3.58
- Win rate: 12.5%