Probabilistic Time Series Forecasting Framework

Develop a sophisticated time series forecasting framework in JavaScript that supports multiple prediction algorithms including ARIMA, exponential smoothing, and Bayesian structural time series models. Create a modular architecture that can handle various temporal data formats, automatically detect seasonality, and generate probabilistic prediction intervals. Implement advanced feature extraction techniques, support for external regressors, and interactive visualization of forecast uncertainties. Include comprehensive error metrics and model performance evaluation.