Skip to Content
avatar image
Former Member

Time Series for beginners

Hello,

I’m trying to use time series. When we have a list of dates and I pick the last date in the list I get the message: “NO future values detected after the last training date given:…”

I saw in the example file that if I would like to predict future values, The key should exist in the file and it will predict it. But if we use a databse taable should we create empty records with the key?

I also noticed that when we don’t have enough data we get the message : Error     There is not enough rows to use this cutting strategy.” Is it possible than in such cases to return an average value or something similar? How can we know how much data is enough?

Thanks,

Amir

Add comment
10|10000 characters needed characters exceeded

  • Get RSS Feed

1 Answer

  • Best Answer
    Sep 06, 2016 at 06:23 AM

    Hi Amir,

    On your first question;

    are you using extra predictable variables when creating your time series model? If that's the case you need to fill the values of these extra predictable variables for the future values. As an example you can refer to the file CashFlows.txt available here in my local installation: C:\Program Files\SAP Predictive Analytics\Desktop 3.0\Automated\Samples\KTS. When opening this file you can notice the values for all variables are filled for the year 2002 but the output.

    For a complete walkthrough you can either refer to the help guide here: http://help.sap.com/businessobject/product_guides/pa30/en/pa30_ts_user_en.pdf or to Andreas Foster tutorial here: http://scn.sap.com/docs/DOC-69324

    On your second question, can you please explain how many data points you are currently using and how many you would like to forecast? What is the level of granularity of your dataset?

    Thanks & regards,

    Antoine

    Add comment
    10|10000 characters needed characters exceeded

    • The two parameters you should look for are the below:

      - t.changeParameter Parameters/LastRow 22 is the last row that contains fully filled training data. It means you have 22 data points for training the model.

      - t.changeParameter Parameters/LastRowWithForecastingInformation 23 is the last row including the point to forecast. This one should be filled with the date at least and extra predictable variables if this applies.

      I hope this clarifies.
      How many data points do you have for learning, how many data points do you want to forecast?

      Best regards

      Antoine