smis: scaled Mean Interval Score (sMIS)

In thfuchs/tsRNN: Time Series Forecasting using Recurrent Neural Networks by Keras

Description

sMIS scaled according to M4 Forecasting competition (see references)

Usage

smis(data, actual, lower, upper, m, level)

Arguments

data	time series (only train set)
actual	actual values (Equivalent of upper/lower)
lower	lower bound of prediction interval
upper	upper bound of prediction interval
m	frequency, e.g. 12 for monthly and 4 for quarterly series
level	level used for prediction interval construction

Value

numeric vector of length 1

References

• Svetunkov, I., Sagaert, Y. R. (2020). greybox: Toolbox for Model Building and Forecasting. https://cran.r-project.org/package=greybox

• Gneiting, T., & Raftery, A. E. (2007). Strictly proper scoring rules, prediction, and estimation. Journal of the American Statistical Association, 102(477), 359–378. https://doi.org/10.1198/016214506000001437

• Makridakis, S., Spiliotis, E., & Assimakopoulos, V. (2020). The M4 Competition: 100,000 time series and 61 forecasting methods. International Journal of Forecasting, 36(1), 54–74. https://doi.org/10.1016/j.ijforecast.2019.04.014

See Also

Other PI accuracy measures: acd()

Examples

data <- tsRNN::fc_arima[key == "actual", value]
forecast <- tsRNN::fc_arima[key == "predict", value]
data_train <- data[1:(length(data) - length(forecast))]
data_test <- data[(length(data) - length(forecast) + 1):length(data)]
lower <- tsRNN::fc_arima[key == "predict", lo95]
upper <- tsRNN::fc_arima[key == "predict", hi95]

tsRNN::smis(
  data = data_train,
  actual = data_test,
  lower = lower,
  upper = upper,
  m = 4,
  level = 0.95
)

thfuchs/tsRNN documentation built on April 17, 2021, 11:03 p.m.