Plotting `forecast` prediction using `dygraphs`

问题

I want to plot a forecast package time series model's predictions using dygraphs. The documentation suggests the following approach for predictions with actuals:

hw <- HoltWinters(ldeaths)
p <- predict(hw, n.ahead = 36, prediction.interval = TRUE)
all <- cbind(ldeaths, p)

dygraph(all, "Deaths from Lung Disease (UK)") %>%
  dySeries("ldeaths", label = "Actual") %>%
  dySeries(c("p.lwr", "p.fit", "p.upr"), label = "Predicted")

Resulting in:

The interesting thing about the plotted object all is its class:

> class(all) [1] "mts" "ts" "matrix"

> is.mts(all)
[1] TRUE
> is.ts(all)
[1] TRUE
> is.matrix(all)
[1] TRUE

str provides a little more information about the object all:

> str(all)
 Time-Series [1:108, 1:4] from 1974 to 1983: 3035 2552 2704 2554 2014 ...
 - attr(*, "dimnames")=List of 2
  ..$ : NULL
  ..$ : chr [1:4] "ldeaths" "p.fit" "p.upr" "p.lwr"

More inspection shows that all is an array:

> tail(all)
         ldeaths     p.fit    p.upr     p.lwr
Jul 1982      NA 1128.3744 1656.127  600.6217
Aug 1982      NA  948.6089 1478.090  419.1282
Sep 1982      NA  960.1201 1491.429  428.8112
Oct 1982      NA 1326.5626 1859.802  793.3235
Nov 1982      NA 1479.0320 2014.306  943.7583
Dec 1982      NA 1929.8349 2467.249 1392.4206
> dim(all)
[1] 108   4
> is.array(all)
[1] TRUE

I am unable to create this type of object using predictions from the forecast package

With my forecast model unemp.mod I create predictions:

> f <- forecast(unemp.mod)
> f
         Point Forecast    Lo 80    Hi 80    Lo 95    Hi 95
Apr 2017       4.528274 4.287324 4.769224 4.159773 4.896775
May 2017       4.515263 4.174337 4.856189 3.993861 5.036664
Jun 2017       4.493887 4.055472 4.932303 3.823389 5.164386
Jul 2017       4.479992 3.936385 5.023599 3.648617 5.311367
Aug 2017       4.463073 3.807275 5.118871 3.460116 5.466030

While it looks similar to the array in the example, it's a totally different object:

> class(f)
[1] "forecast"
> str(f)
List of 10 <truncated>

If I try to generate the forecast using base R's predict like in the example, I also wind up with a list object:

> predict(unemp.mod, n.ahead = 5, prediction.interval = TRUE)
$pred
          Apr      May      Jun      Jul      Aug
2017 4.528274 4.515263 4.493887 4.479992 4.463073

$se
           Apr       May       Jun       Jul       Aug
2017 0.1880140 0.2660260 0.3420974 0.4241788 0.5117221

Does anyone have any suggestions on how to create the right object to plot using dygraphs based on forecast model predictions?

回答1:

Upon further investigation of the list generated by forecast(model) I noticed the actuals and point forecasts are given as ts objects and the upper and lower bounds are in the same array format as the dygraphs HoltWinters example. I created a function that creates the array needed for plotting supposing forecast_obj <- forecast(model).

gen_array <- function(forecast_obj){

  actuals <- forecast_obj$x
  lower <- forecast_obj$lower[,2]
  upper <- forecast_obj$upper[,2]
  point_forecast <- forecast_obj$mean

  cbind(actuals, lower, upper, point_forecast)
}

Note that the lower and upper bounds are 2 dimensional arrays. Since dygraphs does not support more than one prediction interval I only pick one pair (the 95%).

I then plot the resulting array using something like this:

dygraph(ts_array, main = graph_title) %>% 
      dyRangeSelector() %>% 
      dyRangeSelector(height = 40,
                      dateWindow = c("2011-04-01", "2019-4-01")) %>%
      dySeries(name = "actuals", label = "actual") %>%
      dySeries(c("lower","point_forecast","upper"), label = "Predicted") %>%
      dyLegend(show = "always", hideOnMouseOut = FALSE) %>%
      dyHighlight(highlightCircleSize = 5,
                  highlightSeriesOpts = list(strokeWidth = 2)) %>%
      dyOptions(axisLineColor = "navy", gridLineColor = "grey")

Resulting in this graph:

回答2:

In ?dygraphs::dygraph, the data argument has to be

Either time series data or numeric data. For time series, this must be an xts object or an object which is convertible to xts. For numeric data, this must be a named list or data frame, where the first element/column provides x-axis values and all subsequent elements/columns provide one or more series of y-values.

Hence, you will need to extract the data from the forecast class using summary. Convert it into a xts class then use dygraph.

library(forecast)
f <- forecast(ldeaths)

#use summary to get the model fit by forecast package
df <- summary(f)

#convert into a Time-Series class
dfTs <- xts(df, as.Date(paste("01", rownames(df)), "%d %b %Y"))

#plot
dygraph(dfTs)

来源：https://stackoverflow.com/questions/43624634/plotting-forecast-prediction-using-dygraphs

标签

forecasting

dygraphs