Skip to Content
Author's profile photo Tammy Powlas

Looking at my own water consumption with SAP Lumira & SAP Predictive Analytics

I really haven’t paid attention to my own water consumption; out of all the utilities this is the cheapest bill for me.  But for fun I thought I would try to forecast my consumption and compare the predictive capabilities of both Lumira and Predictive Analytics.

/wp-content/uploads/2015/07/1ctrly_742102.jpg

Above is my data acquisition in Excel; consumption is recorded on a quarterly basis.

/wp-content/uploads/2015/07/2formatdate_742160.jpg

I create a date/time hierarchy on the date

/wp-content/uploads/2015/07/3predictive_742161.jpg

Above shows the Predictive options

/wp-content/uploads/2015/07/4consumption_742162.jpg

I’m not sure I agree with the forecast, but it was interesting to see my consumption over the years.  Around 2004-2005 time frame, I had a water garden in my back yard; I didn’t want to keep up with it so I stopped the water from feeding the garden.  Then a few years later I disconnected my dishwasher when it stopped working and I decided not to replace it.  Hence, my water consumption is minimal.

Unlike California, where I live we’ve experienced plenty of rain so there’s been no droughts.  My decrease in consumption more had to do with stopping the water garden and stop using the dishwasher.   I never tied the two events to a decrease in water consumption.

For fun, I thought I would try this in SAP Predictive Analytics.  I didn’t have to do any data preparation, I just loaded the spreadsheet and selected an algorithm:

/wp-content/uploads/2015/07/5atriple_742164.jpg

I tried triple exponential smoothing

/wp-content/uploads/2015/07/5predictive_742165.jpg

Above are the parameters

/wp-content/uploads/2015/07/6tripleexponential_742166.jpg

Above are the results from triple exponential smoothing

I changed the algorithm to R Triple Exponential Smoothing and the results look better to me:

/wp-content/uploads/2015/07/7consump_742167.jpg

I don’t understand the spike in 2018 but the predicted values look more reasonable.

Assigned Tags

      2 Comments
      You must be Logged on to comment or reply to a post.
      Author's profile photo Former Member
      Former Member

      Hi,

      Typically  the confidence on the near term forecast is higher, as compared to a forecast that is generated for a time frame much in the future.

      Now normally one would create the time series model and use it for forecasting on a windowing manner. So as the actual window nears the 2017 timeframe, the forecast will be more accurate for 2018.
      I hope this helps...

      Paul

      Author's profile photo Antoine CHABERT
      Antoine CHABERT

      Hi Tammy Powlas

      I would like to discuss this example with you if possible. Can you please send me an email using my SAP email address?

      thanks

      Antoine