NNMSE: NNMSE - Normalized Naive Mean Squared Error
In jdestefani/UEMTS: Univariate Error Measures for Time Series forecasting

Description Usage Arguments Value Examples

NMSE = 1/n (∑_{t=0}^{n} (y_t-y_hat_t)^2)/(∑_{t=0}^{n} (y_t-y_hat_{Naive,t})^2)

1	NNMSE(y, y_hat, y_hat_naive, result_type = c("measure", "raw"))

`y`	True values of the time series
`y_hat`	Forecasted values of the time series
`y_hat_naive`	Forecasted values of the time series with the naive method
`result_type`	String indicating whether the function should return the raw values ("raw"), or the computed measure ("measure")

According to the value of result_type:

measure -> NNMSE computed according to the formula
raw -> Vector containing one NNMSE for each time step

y <- AirPassengers
splitting_point <- round(2*length(y)/3)
y_train <- y[1:splitting_point]
h <- 5

y_test <- y[(splitting_point+1):(splitting_point+h)]
y_hat_naive <- rep(tail(y_train,1),h)
y_hat_average <-  rep(mean(y_train),h)
results <- NNMSE(y_test,y_hat_average,y_hat_naive)