Bayesforecast 0.0.1 for time series modeling

Publicity
1

I am happy to announce that the beta version of bayesforecast 0.0.1 is already on CRAN.
bayesforecast is a package for Bayesian time series modeling using Stan.

Currently, it supports

  • Basic models such as ARIMA, ARMAX, local level, Holt-trend methods, and lgt models.
  • For seasonality: Sarima, Harmonic Regression, and seasonal Holt-Winters methods.
  • For volatility: GARCH, SVM, and asymmetric GARCH.
  • At least 12 different priors distributions for the model parameters.
  • prophet’s GAM’s models for the automatic forecast.
  • An interface similar to rstanarm and forecast packages.

Special thanks to @avehtari @paul.buerkner and @bgoodri for their help. Any collaboration on documentation, bug detection, and method improvements will be appreciated.

18 Likes
2

Interesting, I will sure keep this in mind when I do temporal modeling in the future! :)

3

Hi Asael,

I see you used to maintain a package called varstan. Do you have any plans to extend bayesforecast to support multivariate analysis?

Thank you!

Sam