AIOps

AIOps

Join this online group to communicate across IBM product users and experts by sharing advice and best practices with peers and staying up to date regarding product enhancements.

 View Only

Performance Analyzer Confidence Values are Greater than 100

By Kristen Meren posted Fri December 01, 2017 08:06 AM

  

by Sandra Jones

This set of blogs will be short entries about issues seen recently and have been solved.  
Just generally anything that might be useful.
Each blog will be on one subject.

Performance analyzer confidence values are greater than 100 within ITPA Non Linear Trending Forecasts (NLT) can be done.
For more information on how it is set up check out the users guide:

http://www.ibm.com/support/knowledgecenter/SSTFXA_6.3.0/com.ibm.itm.doc_6.3/ic/landing_itm_pauser.htm


For these forecast,  historical data from the Tivoli Data Warehouse is used for the trend calculation and is completed with a
nonlinear trend of the predicted forecast.


The strength of the forecast is determined by the number of samples and the confidence.

It was seen on the search results webpage that the confidence value on results was over 100 and sometimes there was even negative numbers in this field.

The value should always be between 0-100.


The issue was discovered to be due to a decimal symbol being a comma ',' rather than a full stop '.' in the warehouse saved data.

It was noted that this worked correctly elsewhere in ITM for the values, but this particular calculation was not using the decimal symbol correctly.  


Therefore if a comma rather than a full stop is used as the decimal symbol, the following steps need to be followed to change the
PA_Fcast_Master.sps. file in ITPA.

1)  stop ITPA
 
2) go into folder

OMS
  /SELECT TABLES
  /IF COMMANDS=[ ' TSModel']
        SUBTYPES = ['ModelStatistics' 'Forecast' 'Model Description']
  /DESTINATION VIEWER=NO FORMAT=TABTEXT
        OUTFILE='tsmodelout.txt'.
 
OMSLOG FILE='omslog.out'
  /APPEND=YES
  /FORMAT=TEXT.
 
PREDICT THRU    DAY 680.
 
TSMODEL
   /MODELSUMMARY  PRINT=[NONE]
   /MODELSTATISTICS  DISPLAY=YES MODELFIT=[ SRSQUARE RSQUARE RMSE MAPE
MAE MAXAPE MAXAE NORMBIC]
   /MODELDETAILS  PRINT=[ FORECASTS]
   /OUTPUTFILTER DISPLAY=ALLMODELS
   /AUXILIARY  CILEVEL=95 MAXACFLAGS=24
   /MISSING USERMISSING=INCLUDE
   /MODEL DEPENDENT=V3
      PREFIX='Model'
   /EXPERTMODELER TYPE=[ARIMA EXSMOOTH] TRYSEASONAL=YES
   /AUTOOUTLIER  DETECT=OFF TYPE=[ ADDITIVE LEVELSHIFT INNOVATIONAL
TRANSIENT LOCALTREND ADDITIVEPATCH].
OMSEND.
RESTORE
------------------------------------------------------------------------
 

5) After making the modification, start ITPA  and run the NLT tasks

Once this was done the confidence values showed as between 0-100.

0 comments
1 view

Permalink