Challenge #24: ARIMA Time Series

[Laszlo_D]

I found that the ARIMA method doesn't calculate with the seasonal effect in December. The results are missleading. I also used Neural Network to predict the units sold based on the week number. Here is my solution:

[aazimh]

My forecast values match the output file, but my confidence interval values don't - not sure why.

[Tomas_Kozlik]

just started with challenges - my solution attached

[justindavis]

Spoiler

[MichaelSu]

Solution Attached. Avg Ignore 0's...nice

[Jonathan-Sherman]

Challenge 24 is done!

Spoiler

[Laszlo_D]

In this update I created an ARIMA forecast calculating with seasonality. This model's predictions are more accurate than the original solution's.

[Laszlo_D]

My improved solution with ARIMA custom Seasonal Components settings.

[KevinHarrison]

My numbers don't match 100% to solution numbers because I filled the weekly nulls then averaged the month per the instructions. Then I created the yearly average with the weeks filled. That makes a little more sense to me but I guess both are making assumptions of some point. 

Spoiler

[Dazzerman]

I spent too long trying to figure out why my confidence values differed from the given solution before I checked to find out that others also had differences...

Anyway, here is my solution.

Spoiler

I see that others haven't included Unions after their Join tools, and I realise that it is possible to avoid them in appropriate ways, but I am in the habit of using them to avoid losing rows so don't usually think about when I don't have to use them. In any case, a fairly simple exercise though had to think a bit about how to use the ARIMA tool as I haven't used it before.