R/forecasting-random-walk.R

In ha0ye/MATSS-forecasting: Analysis and Synthesis of Forecasting Ecological Time Series

#' @title Forecast using a random walk model
#' @aliases randomwalk_one_step
#'
#' @description `randomwalk_ts` fits a random walk model using 
#'   \code{\link[forecast]{rwf}}
#'
#' @inheritParams forecast::rwf
#' @inheritParams forecast_iterated
#' @inheritParams forecast::forecast
#' 
#' @export
#'
randomwalk_ts <- function(timeseries, num_ahead = 5, level = 95,
                          drift = FALSE)
{
    f <- function(training, observed, level, drift)
    {
        # make forecasts
        forecasts <- forecast::rwf(training, NROW(observed), drift = drift, level = level)
        
        # return
        data.frame(observed = as.numeric(observed),
                   predicted = as.numeric(forecasts$mean),
                   lower_CI = as.numeric(forecasts$lower),
                   upper_CI = as.numeric(forecasts$upper))
    }
    
    forecast_iterated(fun = f, timeseries = timeseries, num_ahead = num_ahead,
                             level = level, drift = drift)
}

#' @rdname randomwalk_ts
#' 
#' @description `randomwalk_one_step` uses \code{\link[forecast]{rwf}} to fit a 
#'   random walk model (with or without drift) and make a single one-step 
#'   forecast. 
#'   
#' @export
#'
randomwalk_one_step <- function(timeseries, level = 95, drift = FALSE)
{
    f <- function(training, observed, level, drift)
    {
        # make forecasts
        forecasts <- forecast::rwf(training, 1, drift = drift, level = level)
        
        # return
        data.frame(observed = as.numeric(observed),
                   predicted = as.numeric(forecasts$mean),
                   lower_CI = as.numeric(forecasts$lower),
                   upper_CI = as.numeric(forecasts$upper))
    }
    
    hindcast(fun = f, timeseries = timeseries, 
                      level = level, drift = drift)
}