replicate excel formula functionality in R for time series forecasting

Question

In excel I can have the the following format:

alpha .10

    time   Demand  Forecast
1     10    50        50
2     11    75        
3     12    65
4     13    90
5     14    120

where in the forecast column I would apply simple exponential smoothing to forecast the second period

[(alpha * 75) + (1- alpha)*(50)] 

repeating this process to obtain the additional forecast

[(alpha * next demand) + (1- alpha)*(previous forecast)]

to obtain the 3 - 6th period forecasts.

In R when I try using the forecast package it keeps forecasting for periods 6 and beyond which is not what I want the code I used was:

fore3 <- ses(fore, h = 9, alpha = 0.1)

with dplyr I tried the following:

fore3 <- mutate(fore, Forecast = (alpha * Demand) + lag(Forecast * (1-alpha)))

but end up with

   time   Demand  Forecast
1     10    50        50
2     11    75        NA
3     12    65        NA
4     13    90        NA
5     14    120       NA

Answer 1

Exponentially weighted moving averages are easily implemented with Reduce with acc=TRUE

ewma<-function(x,alpha) Reduce(function(y,xi) y*(1-alpha)+xi*alpha,x,acc=TRUE)

The first argument of Reduce is a function that describes the current step. This should look similar to your function. Specifying acc=TRUE just instructs Reduce to keep the results of each application of the steps. This is analogous to writing a formula in Excel that refers to the previous cell and an adjacent input and dragging the formula down.

ewma(Demand,0.1)
#> [1] 50.0000 52.5000 53.7500 57.3750 63.6375

Answer 2

I'm not entirely sure what you want, but I think you want to apply your function iteratively? How big is your time series? You could use a for loop as follows:

alpha <- 0.1
forecast <- function(x) (alpha * 75) + (1-alpha) * x
periods <- 5
forecast.vector <- 50
for (i in 1:periods) {
    forecast.vector[i + 1] <- forecast(forecast.vector[i])
}
forecast.vector