A solution to last week's Challenge has been posted here. I loved seeing the variety of solutions that our Challengers came up with! Admittedly, my solution probably has more tools than it should but it gets the job done...with only one Join tool! In the spirit of friendly competition, especially with the self-imposed challenge of "how many Join tools does it take to calculate pumpkin production", I thought I'd give a shout out to our "winner". Last week, our Challengers' solutions contained an average of 4 Join tool. Our "winner" with the Fewest Join Tools used: @vishalgupta, with 0 Join tools! You read that right...ZERO! There are a few Find/Replace tools in there, though...but I'll take it! Take the time to check out everyone's solutions; the variety is awesome, and I learned a few new tricks myself!
This week's Challenge will give you the chance to show off your skills with the Predictive Tools to answer one of the most pressing issues of our time: Have we reached Peak Pumpkin? Is the time of the demand for EVERYTHING PUMPKIN losing steam? Has Maple Pecan become a force to be reckoned with?
For this week's Challenge, we have five input files, one for each branch of a store, with data on pumpkin product sales from 2010 to 2016. Amalgamate the data into a single data stream to forecast the total expected sales for each product category for the year 2017.
Notes:
- Any Null values should be filled with the average value of that product for that week, across all stores.
- You will need to install the Predictive tools to use the Time Series tools (if you don't have them installed already).
- Choose the time series model whose Akaike Information Criterion, corrected (abbreviated by AICC) are consistently lower. The data from Store 1 is the best and most complete dataset to use for deciding on a model since it contains no Null values.
- The Predictive District on the Gallery has some tools and samples that you may find helpful for this exercise!
Notes for Users on versions 2018.2 and more recent:
Changes to the version of R used in the Predictive tools have caused ARIMA calculations to yield different results than those in the original post. Please reference the following instructions and start/solution files:
- Any Null values should be filled with the average value of that product for that week, across all stores.
- You will need to install the Predictive tools to use the Time Series tools (if you don't have them installed already).
- Perform this challenge using an ETS model to forecast values for the next year. Leave all settings, aside from those needed to configure the model, to "Auto".
- Expect your output to contain negative values
- Refer to the start file "challenge_94_2018_2_start_file.yxmd" and solution file "challenge_94_2018_2_solution.yxmd" that I posted to my reply to another Community user on page 3 of this Challenge's post.
