How to use a custom function In R Language

Method 1: Using Metrics Package

Another method is to create our own function that calculates the symmetric mean absolute percentage error using the formula that was described before.

Example:

In this program, we have created a custom function computeSmape() that takes actual and forest vectors as input. Using the formula, we are calculating smape. At last, we have called this function by passing vectors as parameters.

R

# Custom function to calculate SMAPE
# for the provided model
computeSmape <- function(actual, forecast) {
  return (1 / length(actual) * sum(2 * abs(forecast - actual) / (abs(actual)  + abs(forecast)) * 100))
}
 
# Defining a vector containing actual values
actual <- c(10, 33, 42, 18, 19, 21, 22)
 
# Defining a vector containing forecasted values
forecast <- c(16, 19, 24, 27, 25, 36, 48)
 
# Determine SMAPE
computeSmape(actual, forecast)

Output:

Thus symmetric mean absolute percentage error for this model comes out to be equal to 49.819% .

How to Calculate SMAPE in R

SMAPE stands for symmetric mean absolute percentage error. It is an accuracy measure and is used to determine the predictive accuracy of models that are based on relative errors. The relative error is computed as:

relative error = x / y
Where x is the absolute error and y is the magnitude of exact value

SMAPE has both lower bound and upper bound values (in contrast to the mean absolute percentage error). SMAPE can be calculated easily using the following formula:

SMAPE = (1 / n) * Σ(|forecast_value – actual_value| / ((|actual_value| + |forecast_value|) / 2) * 100

Here,

Σ : a symbol that means “sum”

|| : represent the absolute value

n : It represents the sample size

actual_value: It represents the actual data value

forecast_value : It represents the forecasted data value

This article focuses on how we can calculate the SMAPE in R: