In davidtedfordholt/fable.bsts: Bayesian Structural Time Series Modelling Interface for 'fable'

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

fable.bsts

This package provides a tidy R interface to the bsts forecasting procedure using fable. This package makes use of the bsts package for R.

While the BSTS() function works in the fable::model() framework, not all functionality from the package has been implemented.

Completed

• Stationary and non-stationary trends
• Regression and trigonometric seasonalities

In Progress

• Holiday models
• Exogenous regressors
• Monthly-annual cycle models
• User-specified priors

Use

# library(tidyverse)
# 
# source("data-raw/lax_passengers.R")
# data <- lax_passengers %>%
#   mutate(date = yearmonth(date))
#   
# data %>%
#   fabletools::model(
#     bsts = BSTS(passengers ~ intercept() + ar() + seasonal("1 day"))
#   ) %>%
#   fabletools::forecast(h = 48) %>%
#   autoplot()

Specials implemented:

• intercept() - static intercept models
• ar() - autoregressive models
• level() - local and shared level models
• trend() - local linear, semilocal linear and Student local linear models
• seasonal() - seasonal regression models
• trig() - trigonometric seasonality models

davidtedfordholt/fable.bsts documentation built on Sept. 15, 2020, 11:38 a.m.