> testData.df
 Date   Sales
1  Jan-13 1004078
2  Feb-13  912734
3  Mar-13  988705
4  Apr-13  902736
5  May-13  939328
6  Jun-13  940007
7  Jul-13  881127
8  Aug-13 1022384
9  Sep-13  919752
10 Oct-13  743247
11 Nov-13 1426428
12 Dec-13  270556
13 Jan-14 1124901
14 Feb-14  896759
15 Mar-14  912444
16 Apr-14 1046760
17 May-14  828060
18 Jun-14 1049858
19 Jul-14 1017681
20 Aug-14 1155440
21 Sep-14 1216358
22 Oct-14 1139405
23 Nov-14 1131746
24 Dec-14 1007372
25 Jan-15  894100
26 Feb-15  753241
27 Mar-15  741080
28 Apr-15  732739
29 May-15  826085
30 Jun-15  813362
31 Jul-15  962813
32 Aug-15  644084
33 Sep-15 1126036
34 Oct-15  889667
35 Nov-15  880956
36 Dec-15  786907
37 Jan-16  706587
38 Feb-16  944412
39 Mar-16  960048
40 Apr-16  878436
41 May-16  784348
42 Jun-16  830803

## ets() forecast with cleaned data
# multicative seasonality (Holy-Winters method)
fit1 <- ets(train_ts.clean, model = "MAM", damped = T)
ets.f1 <- forecast(fit1,
                  #forecast 12 months ahead
                  h = 12,
                  #90% and 99% confidence level
                  level = c(90, 99))
plot(ets.f1)
lines(fit1$states[,1], col = 'red')