Forecasting in R
Rob J. Hyndman
Professor of Statistics at Monash University
Trigonometric terms for seasonality
Box-Cox transformations for heterogeneity
ARMA errors for short-term dynamics
Trend (possibly damped)
Seasonal (including multiple and non-integer periods)
gasoline %>% tbats() %>% forecast() %>%
autoplot() +
xlab("Year") + ylab("thousand barrels per day")
calls %>% window(start = 20) %>%
tbats() %>% forecast() %>%
autoplot() + xlab("Weeks") + ylab("Calls")
Forecasting in R