steve-the-bayesian / BOOM

A C++ library for Bayesian modeling, mainly through Markov chain Monte Carlo, but with a few other methods supported. BOOM = "Bayesian Object Oriented Modeling". It is also the sound your computer makes when it crashes.
GNU Lesser General Public License v2.1
36 stars 15 forks source link

Seasonal models for complex seasonality #2

Open steve-the-bayesian opened 6 years ago

steve-the-bayesian commented 6 years ago

The standard seasonal model in bsts works great, but it assumes changes to seasonal patterns occur as in a random walk. It might be that the seasonal pattern changes predictably throughout the course of the year, and has done so for many years.

To model complex seasonal patterns use the seasonal approach from West and Harrison, where a separate time series model is fit to each season, and then a constraint is imposed to make the seasonal effects sum to zero.

Test this new seasonal model on the turkish electricity data.