plot(DATE, data$inflation, type = 'l')
data$MA4 <- TTR::SMA(data$inflation, n=4)
head(data, n=30)
## # A tibble: 30 x 3
## DATE inflation MA4
## <date> <dbl> <dbl>
## 1 1960-01-01 1.46 NA
## 2 1961-01-01 1.07 NA
## 3 1962-01-01 1.20 NA
## 4 1963-01-01 1.24 1.24
## 5 1964-01-01 1.28 1.20
## 6 1965-01-01 1.59 1.33
## 7 1966-01-01 3.02 1.78
## 8 1967-01-01 2.77 2.16
## 9 1968-01-01 4.27 2.91
## 10 1969-01-01 5.46 3.88
## # … with 20 more rows
library(ggplot2)
pl <- ggplot(data, aes(x = DATE))
pl <- pl + geom_line(aes(y=data$inflation, color = 'inflation'))
pl <- pl + geom_line(aes(y=data$MA4, color = "four period MA"))
pl
## Warning: Removed 3 row(s) containing missing values (geom_path).
ts.data <- ts(data$inflation, start = c(1960,1))
ses <- ses(ts.data, h=2)
summary(ses)
##
## Forecast method: Simple exponential smoothing
##
## Model Information:
## Simple exponential smoothing
##
## Call:
## ses(y = ts.data, h = 2)
##
## Smoothing parameters:
## alpha = 0.9999
##
## Initial states:
## l = 1.4588
##
## sigma: 1.6619
##
## AIC AICc BIC
## 316.7011 317.1221 323.0337
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE
## Training set -0.003691289 1.634429 1.187958 -1.872906 70.54902 0.983622
## ACF1
## Training set 0.1589946
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 2021 1.233642 -0.896168 3.363453 -2.023622 4.490906
## 2022 1.233642 -1.778214 4.245498 -3.372594 5.839879
plot(ses)
hw1 <- holt(ts.data, h=4)
summary(hw1)
##
## Forecast method: Holt's method
##
## Model Information:
## Holt's method
##
## Call:
## holt(y = ts.data, h = 4)
##
## Smoothing parameters:
## alpha = 0.9998
## beta = 1e-04
##
## Initial states:
## l = 0.0362
## b = 0.011
##
## sigma: 1.7013
##
## AIC AICc BIC
## 321.4539 322.5448 332.0083
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE
## Training set 0.008264003 1.644545 1.210022 -0.7917959 72.30846 1.00189
## ACF1
## Training set 0.1537244
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 2021 1.244772 -0.9354931 3.425038 -2.089656 4.579201
## 2022 1.255854 -1.8273884 4.339096 -3.459558 5.971265
## 2023 1.266935 -2.5093315 5.043202 -4.508366 7.042236
## 2024 1.278016 -3.0826005 5.638633 -5.390972 7.947005
plot(hw1)
#5
MSE for Holt Winters = 1.001 MSE for Simple smoothing = 0.98
Eventhough, the MSE for SES is lower than Holt Winters. I will still use Holt Winters for forecasting. Because the forecast values for SES are all the same as the last years value.
On the other hand, Holt Winters prediction are more realistic and hence can be used even with the higher MSE.