Forecasts emphasise probability distributions, expected return ranges and drawdown probabilities. Advanced statistical and ML models will connect via the ModelService abstraction to a separate Python service.
Forecast Card
missing- Expected return, P(positive), lower / median / upper
- Expected volatility, P(drawdown > 10% / 20%)
- Model version, training and OOS periods, confidence
Widget scaffolded. Data provider not yet connected — no synthetic values shown to avoid mixing demo data with production.
Baseline Models
missing- Naive, historical mean, moving average
- Exponential smoothing, linear regression, ARIMA
- Always compared to advanced models
Widget scaffolded. Data provider not yet connected — no synthetic values shown to avoid mixing demo data with production.
Statistical & ML Models
missing- Elastic Net, Random Forest, Gradient Boosting
- XGBoost, LightGBM, SVR
- GARCH, VAR, HMM, Bayesian structural TS
- LSTM / TFT reserved as future extensions
Widget scaffolded. Data provider not yet connected — no synthetic values shown to avoid mixing demo data with production.
Validation
missing- Expanding / rolling / walk-forward windows
- Strict chronological splits, no random shuffles
- MAE, RMSE, directional accuracy, Brier, calibration
- Sharpe, Sortino, max drawdown, profit factor
Widget scaffolded. Data provider not yet connected — no synthetic values shown to avoid mixing demo data with production.