## Data Set
df <- read.csv("//Users//kevinclifford//Downloads//Alcohol_Sales.csv", header=TRUE)

df$Sales <- df$S4248SM144NCEN
df$S4248SM144NCEN <- NULL

ts <- ts(df$Sales, frequency = 12, start=c(1992))

plot(ts)

## ETS Models
fit1 <- ets(ts)
fit1
## ETS(M,Ad,M) 
## 
## Call:
##  ets(y = ts) 
## 
##   Smoothing parameters:
##     alpha = 0.0805 
##     beta  = 0.0232 
##     gamma = 1e-04 
##     phi   = 0.9592 
## 
##   Initial states:
##     l = 4199.083 
##     b = 3.4466 
##     s = 1.1642 1.0362 1.0338 0.9829 1.0534 1.0081
##            1.106 1.0665 0.9754 0.9758 0.8275 0.7702
## 
##   sigma:  0.0455
## 
##      AIC     AICc      BIC 
## 5672.314 5674.549 5740.422
plot(fit1)

accuracy(fit1)
##                    ME     RMSE      MAE       MPE     MAPE      MASE       ACF1
## Training set 53.45892 364.3906 285.5164 0.5004852 3.657119 0.6753521 -0.2918586
fe <- forecast(fit1, 12)
acc <- accuracy(fe, df$Sales[1:12])
acc
##                       ME      RMSE       MAE          MPE       MAPE      MASE
## Training set    53.45892  364.3906  285.5164    0.5004852   3.657119  0.312568
## Test set     -9484.70207 9600.8980 9484.7021 -228.2510458 228.251046 10.383342
##                    ACF1
## Training set -0.2918586
## Test set             NA
plot(fe, main="MMN")

arima1 <- auto.arima(ts)
arima1
## Series: ts 
## ARIMA(3,1,1)(0,1,2)[12] 
## 
## Coefficients:
##           ar1     ar2     ar3      ma1     sma1     sma2
##       -0.1428  0.1580  0.5125  -0.9483  -0.2601  -0.2642
## s.e.   0.0637  0.0651  0.0609   0.0328   0.0581   0.0543
## 
## sigma^2 = 102379:  log likelihood = -2242.28
## AIC=4498.56   AICc=4498.93   BIC=4524.77
plot(arima1)

accuracy(arima1)
##                    ME     RMSE      MAE       MPE     MAPE      MASE       ACF1
## Training set 34.22564 310.4741 232.8522 0.3437269 2.939946 0.5507817 0.02751723
fe2 <- forecast(arima1, 12)
acc2 <- accuracy(fe2, df$Sales[1:12])
acc2
##                       ME      RMSE       MAE          MPE       MAPE       MASE
## Training set    34.22564  310.4741  232.8522    0.3437269   2.939946  0.2549141
## Test set     -9644.03404 9745.1509 9644.0340 -232.2911945 232.291195 10.5577699
##                    ACF1
## Training set 0.02751723
## Test set             NA
plot(fe2, main="Auto-ARIMA")

##Neural networks
time <- fe2$residuals %>% as_tsibble()
beer <- ts %>% as_tsibble
fit <- time %>% model(NNETAR(value))
fit %>%
  forecast(h = 30) %>%
  autoplot(time) +
  labs(x = "Month", y = "Counts", title = "Monthly Beer Sales")

fit3 <- beer %>% model(NNETAR(value))
fit3 %>%
  forecast(h = 30) %>%
  autoplot(beer) +
  labs(x = "Month", y = "Counts", title = "Monthly Beer Sales")

accuracy(fit3)
## # A tibble: 1 × 10
##   .model        .type       ME  RMSE   MAE     MPE  MAPE  MASE RMSSE   ACF1
##   <chr>         <chr>    <dbl> <dbl> <dbl>   <dbl> <dbl> <dbl> <dbl>  <dbl>
## 1 NNETAR(value) Training 0.706  136.  107. -0.0560  1.50 0.252 0.253 0.0640
gg_tsresiduals(fit3)
## Warning: Removed 15 row(s) containing missing values (geom_path).
## Warning: Removed 15 rows containing missing values (geom_point).
## Warning: Removed 15 rows containing non-finite values (stat_bin).

#Neural Network example sunspots <- sunspot.year %>% as_tsibble() fit <- sunspots %>% model(NNETAR(sqrt(value))) fit %>% forecast(h = 30) %>% autoplot(sunspots) + labs(x = “Year”, y = “Counts”, title = “Yearly sunspots”)