R - predict on the base of trend and seasonality

Vote count:

1

I'm writing function used for forecasting sales on the base of trend and seasonality. I use dummy variables to represent seasonality and time variable for trend. Here is the code:

forecast<-function(data, time, fn) {

n <- length(data)
seasonal <- factor(cycle(data))
new_data <- seq(from=cycle(data)[n]+1, length=fn, by=1) 
new_seasonal <- factor(new_data)
trend <- rep(0, n)
new_trend <- rep(0, n)
if (time > 0) {
    trend <- vector()
    new_trend <- vector()

    ttrend <- seq(from=1, to=n, by=1)
    tnew_trend <- seq(from=n+1, length=fn, by=1)

    for(i in 1:time) {
        trend <- cbind(trend, ttrend^i)
        new_trend <- cbind(new_trend, tnew_trend^i)
    }

    model_trend <- lm(data ~ seasonal + trend)
} else {

    model_trend <- lm(data ~ seasonal)
}


df <- data.frame(new_seasonal, new_trend)

    p <- predict(model_trend, df)

}

forecast(data = dane.ts[,"SALES"], time=2, fn=5)

However I get warning:

Warning message: 'newdata' had 5 rows but variable(s) found have 104 rows

And it seems that sth is wrong with this function. I would appreciate any help.

asked 3 mins ago

user2674174

15