modeltime_wfs_forecast: Modeltime workflow sets forecast

In sknifedatar: Swiss Knife of Data

Description

forecast from a set of recipes and models trained by modeltime_wfs_fit() function.

Usage

modeltime_wfs_forecast(.wfs_results, .series, .split_prop = NULL, .h = NULL)

Arguments

.wfs_results	tibble of combination of recipes and models fitted, generated with the modeltime_wfs_fit() function.
.series	time series dataframe.
.split_prop	time series split proportion.
.h	time series horizon from the modeltime_forecast() function from 'modeltime' package.

Details

since it uses the modeltime_forecast() function from 'modeltime' package, either the forecast can be made on new data or on a number of periods.

Value

a tibble containing the forecast for each model.

Examples

library(dplyr)
library(modeltime)
library(earth)

data <- sknifedatar::data_avellaneda %>%  mutate(date=as.Date(date)) %>% 
filter(date<'2012-06-01')

recipe_date <- recipes::recipe(value ~ ., data = data) %>% 
  recipes::step_date(date, features = c('dow','doy','week','month','year')) 

mars <- parsnip::mars(mode = 'regression') %>%
  parsnip::set_engine('earth')

wfsets <- workflowsets::workflow_set(
  preproc = list(
    R_date = recipe_date),
  models  = list(M_mars = mars),
  cross   = TRUE)

wffits <- sknifedatar::modeltime_wfs_fit(.wfsets = wfsets, 
                                         .split_prop = 0.8, 
                                         .serie=data)
                            
sknifedatar::modeltime_wfs_forecast(.wfs_results=wffits, 
                                    .series = data,
                                    .split_prop = 0.8) 
                                     

sknifedatar documentation built on June 1, 2021, 9:08 a.m.