"Prediction steps" has no effect for "Microsoft Time Series" algorithm

Hot Topic

Converting T-SQL *= OUTER JOINS to ANSI-92 syntax
"Open Table" NOT WORKING. Can someone HELP!
How large is 4 GB
PIVOT statement whitout knowing values
How do you find out whether PC is 32bit or 64bit
Interactive charts, etc.
Get back data from Reports
Master-detail on the same dataset
Did SP1 fix Query Optimizer issues w/ Partitioned tables?
Accessing the Report Builder Application

SQL Server

How to initialize pull transactional subscription with attaching of a database
XML Service Broker messages
default schema for NT-group not possible
How to set a ReportParameter to the value NULL?
Sql 2005 SP1 (x86) upgrade failed....
Member count limit
dynamically sorting, filtering, and grouping ????
NullProcessing and Reference Dimensions
Importing unicode data from Oracle Database
SQL Server Reporting Services Installation - Urgent

"Prediction steps" has no effect for "Microsoft Time Series" algorithm

I am using "Microsoft Time Series" algorithm.  When I view the charts under "Mining Model Viewer", I see only the predicted points of only one step.  When I change the "Prediction steps" value, the scale of the time axis extends, but the actual curve does not change at all (i.e. no points after one step).  I must be missing something rudimentary. Could anyone help

Thanks,

H




Answer this question

"Prediction steps" has no effect for "Microsoft Time Series" algorithm

  • Subhash Bhave

    Thanks! BTW, I am using real data, not the sample, doing meaningful work.

    I have another question: how can I get historical values in my prediction query. For example, suppose I am predicting X(1), (2) based on X(0), X(-1), .... I can get X(1) by functions such as Predict(X). How can I get X(0) or X(-1) in the same DMX query. The DM studio's chart does exactly this. I hate to waste anyone's time for such trivial questions. If I can be pointed to source for answering such questions, it would be great.

    I am enjoying reading Data Mining with SQL Server 2005 that is quite helpful.



  • Yura Developer

    This can happen if the algorithm detects that it cannot generate an accurate prediction.

    In the case the algorithm detects that predicted values will explode expotentially, it returns NULL values.



  • Kmicic77

    Is there a place (e.g. a log file) that I can check to confirm this is what happens.

    After exploring the SQL DM for a few days with the help of online book, I have given up and am waiting for the arrival of your book Data Mining with SQL Server 2005 before I resume my exploration though I am eager to start the development of my application.  I have too many questions to be posted here. Usually online material is sufficient for me to learn new features and use them, but the SQL 2005 online book is way too sketchy for the DM part unless I miss some part with detailed description of DM.

    Thanks

     



  • Sunny9001

    Not really - that's just what happens. If you use the sample AdventureWorks time series model you will see that various series predict out to various lengths.

    If you are asking for more predictions than the model can provide, it will return NULL. If there is an error - it will return an error.



  • "Prediction steps" has no effect for "Microsoft Time Series" algorithm
Answer this question