In markfairbanks/parable: Parallel tools for 'fable'

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

parable

The goal of parable is to enable the user to use fable in parallel.

Installation

You can install the development version from GitHub with:

# install.packages("devtools")
devtools::install_github("markfairbanks/parable")

Note that parable uses a custom version of fabletools, which can be installed here:

# install.packages("devtools")
devtools::install_github("markfairbanks/fabletools")

Functions

parable has 3 core functions:

• parallel_model()
• parallel_forecast()
• parallel_accuracy()

Each function is used just like the functions from fable, but a future::plan must be set.

We start with an example tsibble with 50 time-series.

pacman::p_load(tidyverse, parable, janitor, lubridate,
               tsibble, fable, fabletools, feasts, tsibbledata)

aus_ts <- tsibbledata::aus_retail %>%
  rename_all(janitor::make_clean_names) %>%
  update_tsibble(key = series_id, index = month) %>%
  select(-state, -industry) %>%
  filter(year(month) >= 2010) %>%
  group_by_key() %>%
  filter(group_indices() <= 50) %>%
  ungroup()

aus_ts

The below chunk shows how to use parable, with tictoc used to show the timing of the parallel_model() step:

pacman::p_load(future)

plan(multiprocess, workers = 5)

tictoc::tic()

parable_mbl <- aus_ts %>%
  parallel_model(ets = ETS(turnover))

tictoc::toc()

Which we can then compare to the timing of normal fable:

tictoc::tic()

fable_mbl <- aus_ts %>%
  model(ets = ETS(turnover))

tictoc::toc()

markfairbanks/parable documentation built on Feb. 24, 2020, 1:58 a.m.