Skip to content

Instantly share code, notes, and snippets.

@ha0ye
Created March 12, 2020 16:35
Show Gist options
  • Save ha0ye/d62b33b1d735b09610852a7113f6bda3 to your computer and use it in GitHub Desktop.
Save ha0ye/d62b33b1d735b09610852a7113f6bda3 to your computer and use it in GitHub Desktop.
Forward forecasts using rEDM
library(rEDM)
#### Example 1: a single forward prediction
# Assume we have already done the analysis, finding that the best embedding
# dimension is 2. Our objective is to make multiple 1-step ahead forecasts
# using simplex and E = 2.
dat <- data.frame(yr = as.numeric(time(sunspot.year)),
sunspot_count = as.numeric(sunspot.year))
E <- 3
# setup data structures:
# 1) add empty row with incremented time, and empty variable of interest
dat <- rbind(dat,
data.frame(yr = tail(dat, 1)$yr + 1,
sunspot_count = NA))
# 2) setup lib and pred:
# lib excludes the last row;
# pred includes the last row and E previous lags
lib <- c(1, NROW(dat) - 1)
pred <- c(NROW(dat) - E, NROW(dat))
# 3) make prediction
simplex_output <- simplex(dat, lib = lib, pred = pred, E = E,
stats_only = FALSE)
# 4) replace the missing value in `dat` with the forecast, so that we can
# repeat the process
dat[NROW(dat), "sunspot_count"] <- simplex_output$model_output[[1]][E, "pred"]
#### Example 2: multiple forward predictions
# setup initial data and E as previously
dat <- data.frame(yr = as.numeric(time(sunspot.year)),
sunspot_count = as.numeric(sunspot.year))
E <- 3
# wrap up previous steps 1-4 in a loop
num_predictions <- 5
for (i in seq(num_predictions))
{
dat <- rbind(dat,
data.frame(yr = tail(dat, 1)$yr + 1,
sunspot_count = NA))
lib <- c(1, NROW(dat) - 1)
pred <- c(NROW(dat) - E, NROW(dat))
simplex_output <- simplex(dat, lib = lib, pred = pred, E = E,
stats_only = FALSE)
dat[NROW(dat), "sunspot_count"] <- simplex_output$model_output[[1]][E, "pred"]
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment