Explorer Forecast: Can I see the transformed dataset fed to Base Learner?

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

Explorer Forecast: Can I see the transformed dataset fed to Base Learner?

Chaitanya Birudavolu
Hi,

In the Forecast Panel of Weka's Explorer, is it possible to see the final transformed dateset (including the added lag variables and date-derived variables), that is being fed to the chosen Base Learner algorithm?

I believe it would help me immensely in understanding how the various configuration settings work, including how the Skip List works.

I''m using 3.8.1, and have already read Mark Hall's well-written wiki "Time Series Analysis and Forecasting with Weka" (https://wiki.pentaho.com/display/DATAMINING/Time+Series+Analysis+and+Forecasting+with+Weka), but many questions still linger.

Thanks and Regards,
Chaitanya
_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|

Re: Explorer Forecast: Can I see the transformed dataset fed to Base Learner?

Mark Hall
Yes you can see what the transformed data looks like. In recent versions of the timeserieseForecasting package, the data transformations have been encapsulated in a Weka filter. This is available in the Explorer, KnowledgeFlow and on the command line. E.g.

java weka.Run .TSLagMaker -h

Will print the help and options for the lag maker.

Cheers,
Mark.

On 16/11/18, 5:24 PM, "Chaitanya Birudavolu" <[hidden email] on behalf of [hidden email]> wrote:

    Hi,
   
    In the Forecast Panel of Weka's Explorer, is it possible to see the final transformed dateset (including the added lag variables and date-derived variables), that is being fed to the chosen Base Learner algorithm?
   
    I believe it would help me immensely in understanding how the various configuration settings work, including how the Skip List works.
   
    I''m using 3.8.1, and have already read Mark Hall's well-written wiki "Time Series Analysis and Forecasting with Weka" (https://wiki.pentaho.com/display/DATAMINING/Time+Series+Analysis+and+Forecasting+with+Weka), but many questions still linger.
   
    Thanks and Regards,
    Chaitanya
    _______________________________________________
    Wekalist mailing list
    Send posts to: [hidden email]
    To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
    List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
   


_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|

Re: Explorer Forecast: Can I see the transformed dataset fed to Base Learner?

Chaitanya Birudavolu
In reply to this post by Chaitanya Birudavolu
Hi,

Thanks for the update.

TSLagMaker seems to fit the bill.

Well, almost...

Is there anything like a TSCustomDateDerivedAttribMaker ? (An unsupervised attrib filter that adds custom date-derived attributes, the same way as in Explorer==>Forecast==>PeriodicAttributes==>New)

That's the only part of the dataset transformation missed out by TSLagMaker, I guess.

Thanks and Regards,
Chaitanya

--------------------------------------------
On Wed, 11/21/18, Mark Hall <[hidden email]> wrote:

 Subject: Re: [Wekalist] Explorer Forecast: Can I see the transformed dataset fed to Base Learner?
 To: "Chaitanya Birudavolu" <[hidden email]>, "Weka machine learning workbench list." <[hidden email]>
 Date: Wednesday, November 21, 2018, 9:48 PM
 
 Yes you can see what the
 transformed data looks like. In recent versions of the
 timeserieseForecasting package, the data transformations
 have been encapsulated in a Weka filter. This is available
 in the Explorer, KnowledgeFlow and on the command line.
 E.g.
 
 java weka.Run
 .TSLagMaker -h
 
 Will print
 the help and options for the lag maker.
 
 Cheers,
 Mark.
 
 On 16/11/18, 5:24 PM, "Chaitanya
 Birudavolu" <[hidden email]
 on behalf of [hidden email]>
 wrote:
 
     Hi,
    
     In the Forecast
 Panel of Weka's Explorer, is it possible to see the
 final transformed dateset (including the added lag variables
 and date-derived variables), that is being fed to the chosen
 Base Learner algorithm?
    
     I believe it would help me immensely in
 understanding how the various configuration settings work,
 including how the Skip List works.
    
     I''m using 3.8.1, and have
 already read Mark Hall's well-written wiki "Time
 Series Analysis and Forecasting with Weka" (https://wiki.pentaho.com/display/DATAMINING/Time+Series+Analysis+and+Forecasting+with+Weka),
 but many questions still linger.
    
     Thanks and Regards,
    
 Chaitanya
    
 _______________________________________________
     Wekalist mailing list
  
   Send posts to: [hidden email]
     To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
     List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
    
 
 
 
_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|

Re: Explorer Forecast: Can I see the transformed dataset fed to Base Learner?

Chaitanya Birudavolu
In reply to this post by Chaitanya Birudavolu

In continuation, if there were a TSDateDerivedAttribMaker:

--1--Some of the date-derived attrib creation that logically isn't really "lag-making" could be MOVED from TSLagMaker to TSDateDerivedAttribMaker. For example, TSLagMaker, currently has fields like "Add AM indicator", "Add day of the week", etc, that do not, in a strict technical sense, belong to "lag-making"

--2--The TSDateDerivedAttribMaker could also offer the ability to add custom date-derived attributes (same as in Explorer==>Forecast==>PeriodicAttributes==>New/Edit)

Cheers,
Chaitanya

--------------------------------------------
On Wed, 11/21/18, Chaitanya Birudavolu <[hidden email]> wrote:

 Subject: Re: [Wekalist] Explorer Forecast: Can I see the transformed dataset fed to Base Learner?
 To: "Weka machine learning workbench list." <[hidden email]>, "Mark Hall" <[hidden email]>
 Date: Wednesday, November 21, 2018, 11:07 PM
 
 Hi,
 
 Thanks for the update.
 
 TSLagMaker seems to fit the bill.
 
 Well, almost...
 
 Is there anything like a
 TSCustomDateDerivedAttribMaker ? (An unsupervised attrib
 filter that adds custom date-derived attributes, the same
 way as in
 Explorer==>Forecast==>PeriodicAttributes==>New)
 
 That's the only part of the dataset
 transformation missed out by TSLagMaker, I guess.
 
 Thanks and Regards,
 Chaitanya
 
 --------------------------------------------
 On Wed, 11/21/18, Mark Hall <[hidden email]>
 wrote:
 
  Subject: Re: [Wekalist] Explorer
 Forecast: Can I see the transformed dataset fed to Base
 Learner?
  To: "Chaitanya Birudavolu" <[hidden email]>,
 "Weka machine learning workbench list." <[hidden email]>
  Date: Wednesday, November 21, 2018,
 9:48 PM
 
  Yes you can see what the
  transformed data looks like. In recent
 versions of the
  timeserieseForecasting package, the
 data transformations
  have been encapsulated in a Weka
 filter. This is available
  in the Explorer, KnowledgeFlow and on
 the command line.
  E.g.
 
  java weka.Run
  .TSLagMaker -h
 
  Will print
  the help and options for the lag
 maker.
 
  Cheers,
  Mark.
 
  On 16/11/18, 5:24 PM, "Chaitanya
  Birudavolu" <[hidden email]
  on behalf of [hidden email]>
  wrote:
 
      Hi,
     
      In the Forecast
  Panel of Weka's Explorer, is it
 possible to see the
  final transformed dateset (including
 the added lag variables
  and date-derived variables), that is
 being fed to the chosen
  Base Learner algorithm?
     
      I believe it would help me
 immensely in
  understanding how the various
 configuration settings work,
  including how the Skip List works.
     
      I''m using 3.8.1, and have
  already read Mark Hall's well-written
 wiki "Time
  Series Analysis and Forecasting with
 Weka" (https://wiki.pentaho.com/display/DATAMINING/Time+Series+Analysis+and+Forecasting+with+Weka),
  but many questions still linger.
     
      Thanks and Regards,
     
  Chaitanya
     
 
 _______________________________________________
      Wekalist mailing list
   
    Send posts to: [hidden email]
      To subscribe, unsubscribe, etc.,
 visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
      List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
     
 
 
 
_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|

Re: Explorer Forecast: Can I see the transformed dataset fed to Base Learner?

Mark Hall
Hmm. Does the -custom option for TSLagMaker not work? It does look like there is a problem in the GUI with this filter at the moment though :-) I'll take a look at that.

Yes, I agree that the date-derived attribute creation should really be in a separate filter. Unfortunately, I was well and truly time series'd to death by the time I had the system running. Moving the data transformation logic into a filter implementation was the last gasp on this front :-)

Cheers,
Mark.

On 22/11/18, 12:34 PM, "Chaitanya Birudavolu" <[hidden email]> wrote:

   
    In continuation, if there were a TSDateDerivedAttribMaker:
   
    --1--Some of the date-derived attrib creation that logically isn't really "lag-making" could be MOVED from TSLagMaker to TSDateDerivedAttribMaker. For example, TSLagMaker, currently has fields like "Add AM indicator", "Add day of the week", etc, that do not, in a strict technical sense, belong to "lag-making"
   
    --2--The TSDateDerivedAttribMaker could also offer the ability to add custom date-derived attributes (same as in Explorer==>Forecast==>PeriodicAttributes==>New/Edit)
   
    Cheers,
    Chaitanya
   
    --------------------------------------------
    On Wed, 11/21/18, Chaitanya Birudavolu <[hidden email]> wrote:
   
     Subject: Re: [Wekalist] Explorer Forecast: Can I see the transformed dataset fed to Base Learner?
     To: "Weka machine learning workbench list." <[hidden email]>, "Mark Hall" <[hidden email]>
     Date: Wednesday, November 21, 2018, 11:07 PM
     
     Hi,
     
     Thanks for the update.
     
     TSLagMaker seems to fit the bill.
     
     Well, almost...
     
     Is there anything like a
     TSCustomDateDerivedAttribMaker ? (An unsupervised attrib
     filter that adds custom date-derived attributes, the same
     way as in
     Explorer==>Forecast==>PeriodicAttributes==>New)
     
     That's the only part of the dataset
     transformation missed out by TSLagMaker, I guess.
     
     Thanks and Regards,
     Chaitanya
     
     --------------------------------------------
     On Wed, 11/21/18, Mark Hall <[hidden email]>
     wrote:
     
      Subject: Re: [Wekalist] Explorer
     Forecast: Can I see the transformed dataset fed to Base
     Learner?
      To: "Chaitanya Birudavolu" <[hidden email]>,
     "Weka machine learning workbench list." <[hidden email]>
      Date: Wednesday, November 21, 2018,
     9:48 PM
     
      Yes you can see what the
      transformed data looks like. In recent
     versions of the
      timeserieseForecasting package, the
     data transformations
      have been encapsulated in a Weka
     filter. This is available
      in the Explorer, KnowledgeFlow and on
     the command line.
      E.g.
     
      java weka.Run
      .TSLagMaker -h
     
      Will print
      the help and options for the lag
     maker.
     
      Cheers,
      Mark.
     
      On 16/11/18, 5:24 PM, "Chaitanya
      Birudavolu" <[hidden email]
      on behalf of [hidden email]>
      wrote:
     
          Hi,
         
          In the Forecast
      Panel of Weka's Explorer, is it
     possible to see the
      final transformed dateset (including
     the added lag variables
      and date-derived variables), that is
     being fed to the chosen
      Base Learner algorithm?
         
          I believe it would help me
     immensely in
      understanding how the various
     configuration settings work,
      including how the Skip List works.
         
          I''m using 3.8.1, and have
      already read Mark Hall's well-written
     wiki "Time
      Series Analysis and Forecasting with
     Weka" (https://wiki.pentaho.com/display/DATAMINING/Time+Series+Analysis+and+Forecasting+with+Weka),
      but many questions still linger.
         
          Thanks and Regards,
         
      Chaitanya
         
     
     _______________________________________________
          Wekalist mailing list
       
        Send posts to: [hidden email]
          To subscribe, unsubscribe, etc.,
     visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
          List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
         
     
     
     
   


_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html