build_models: Forecasting models for 'dockless_dfc' object

In luukvdmeer/dockless: Spatio-Temporal Forecasts for Bike Availability in Dockless Bike Sharing Systems

Description

Fits a forecasting model on each dockless_df time series in a dockless_dfc. For non-seasonal time series, an ARIMA model will be fitted on the log transformed data. For seasonal time series, the log transformed data will be decomposed with STL decomposition, and an ARIMA model will be fitted on the seasonally adjusted data.

Usage

build_models(data, auto_seasonality = TRUE, seasons = list(NULL, 96,
  672, c(96, 672)))

Arguments

data	object of class dockless_dfc.
auto_seasonality	logical. If TRUE, the seasonal period of each time series will be automatically determined with time series cross-validation.
seasons	if auto_seasonality is set to TRUE: a list of different seasonal period lengths - or, in the case of multiple seasonality, combinations of seasonal period lenghts - in time lags to be tested for in the time series cross-validation. No seasonality should be represented by NULL. If auto_seasonality is set to FALSE: a list of seasonal period lengths - or, in the case of multiple seasonality, combinations of seasonal period lenghts - in time lags to be used for each of the time series. No seasonality should be represented by NULL. The list should be of the same length as the number of time series in the dockless_dfc object.

Value

Returns a list of models, one for each dockless_df time series. Each model is an object of class ARIMA for non-seasonal data and an object of class stlm for seasonal data. Both classes come from the forecast package.

luukvdmeer/dockless documentation built on May 10, 2019, 1:24 p.m.