Don't show a specific attribute value

[Andreza Vieira]

Hello,

Please, help me.
I want to know if there's a way to generate the rules (Weka - Association Rules) without a specific attribute value. For example: I have an attribute "play" that can assume "true" or "false", so, I want to generate rules that contain the attribute "play" with only the "true" value, is not interesting for me see this attribute with the "false" value.

How can I do it?

I want to do this because I have 50 boolean attributes. So, the Weka generates a lot of rules, with several combinations between them, that is, a lot of not interesting rules.

Thanks a lot.
Andreza

[Mark Hall-9]

On 23/06/2009, at 1:20 PM, Andreza Vieira wrote:

>
> Hello,
>
> Please, help me.
> I want to know if there's a way to generate the rules (Weka -  
> Association
> Rules) without a specific attribute value. For example: I have an  
> attribute
> "play" that can assume "true" or "false", so, I want to generate  
> rules that
> contain the attribute "play" with only the "true" value, is not  
> interesting
> for me see this attribute with the "false" value.
>
> How can I do it?
>
> I want to do this because I have 50 boolean attributes. So, the Weka
> generates a lot of rules, with several combinations between them,  
> that is, a
> lot of not interesting rules.

When processing market basket-type data with Apriori, you need to  
encode the data in a specific way to avoid the "uninteresting" rules  
that contain the false values. There are several ways to accomplish  
this:

1) Standard ARFF format: use single valued attributes (i.e. @attribute  
myAtt {true}) and use missing values to indicate that an item is not  
present (i.e. missing value instead of "false").

2) In the latest version of Weka with standard ARFF format: use  
boolean attributes but make sure that "false" is the first value  
declared for each attribute (i.e. @attribute myAtt {false, true}) and  
use the -Z option (Treat zero (i.e. first value of nominal attributes)  
as missing) with Apriori.

With option 2 you can also use Weka's sparse instances format in order  
to save memory. The algorithm runs fractionaly slower when processing  
sparse instances.

Cheers,
Mark.

--
Mark Hall
Senior Developer/Consultant, Pentaho Open Source Business Intelligence
Citadel International, Suite 340, 5950 Hazeltine National Dr.,
Orlando, FL 32822, USA
[phone] office, [phone] mobile, [phone] fax,
Skype: mark.andrew.hall, Yahoo: mark_andrew_hall
Download the latest release today <http://www.sourceforge.net/projects/pentaho 
 >




_______________________________________________
Wekalist mailing list
Send posts to: Wekalist@...
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

[Andreza Vieira]

In the option 2, what is the Apriori parameter that I have to set 0? I don't know where is the -Z option =(
I'm using the Weka version 3.6.0, is it the latest version?

Thanks,
Andreza

Mark Hall-9 wrote:

On 23/06/2009, at 1:20 PM, Andreza Vieira wrote:

>
> Hello,
>
> Please, help me.
> I want to know if there's a way to generate the rules (Weka -  
> Association
> Rules) without a specific attribute value. For example: I have an  
> attribute
> "play" that can assume "true" or "false", so, I want to generate  
> rules that
> contain the attribute "play" with only the "true" value, is not  
> interesting
> for me see this attribute with the "false" value.
>
> How can I do it?
>
> I want to do this because I have 50 boolean attributes. So, the Weka
> generates a lot of rules, with several combinations between them,  
> that is, a
> lot of not interesting rules.

When processing market basket-type data with Apriori, you need to  
encode the data in a specific way to avoid the "uninteresting" rules  
that contain the false values. There are several ways to accomplish  
this:

1) Standard ARFF format: use single valued attributes (i.e. @attribute  
myAtt {true}) and use missing values to indicate that an item is not  
present (i.e. missing value instead of "false").

2) In the latest version of Weka with standard ARFF format: use  
boolean attributes but make sure that "false" is the first value  
declared for each attribute (i.e. @attribute myAtt {false, true}) and  
use the -Z option (Treat zero (i.e. first value of nominal attributes)  
as missing) with Apriori.

With option 2 you can also use Weka's sparse instances format in order  
to save memory. The algorithm runs fractionaly slower when processing  
sparse instances.

Cheers,
Mark.

--
Mark Hall
Senior Developer/Consultant, Pentaho Open Source Business Intelligence
Citadel International, Suite 340, 5950 Hazeltine National Dr.,
Orlando, FL 32822, USA
[phone] office, [phone] mobile, [phone] fax,
Skype: mark.andrew.hall, Yahoo: mark_andrew_hall
Download the latest release today <http://www.sourceforge.net/projects/pentaho 
 >




_______________________________________________
Wekalist mailing list
Send posts to: [e-mail]
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

[Peter Reutemann-3]

Please no top-posting, see mailing list etiquette why
(http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html).

> In the option 2, what is the Apriori parameter that I have to set 0? I don't
> know where is the -Z option =(

The property in the GUI is called "treatZeroAsMissing".

> I'm using the Weka version 3.6.0, is it the latest version?

No, 3.6.1 is the latest at the moment for the 3.6.x branch (just check
out the Weka homepage).

Cheers, Peter
--
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. [phone]

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist@...
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

[Andreza Vieira]

Peter Reutemann-3 wrote:

Please no top-posting, see mailing list etiquette why
(http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html).

> In the option 2, what is the Apriori parameter that I have to set 0? I don't
> know where is the -Z option =(

The property in the GUI is called "treatZeroAsMissing".

Sorry, I have installed the 3.6.1 version, but I'm not finding the "treatZeroAsMissing" in the Apriori algorithm parameters. Where can I find it?
Thanks a lot.




> I'm using the Weka version 3.6.0, is it the latest version?

No, 3.6.1 is the latest at the moment for the 3.6.x branch (just check
out the Weka homepage).

Cheers, Peter
--
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. [phone]

_______________________________________________
Wekalist mailing list
Send posts to: [e-mail]
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

[Peter Reutemann-3]

> > Sorry, I have installed the 3.6.1 version, but I'm not finding the
> > "treatZeroAsMissing" in the Apriori algorithm parameters. Where can I find
> > it?

Hmm... Quite possible that it's only available in the developer
version (I never used any other version). The latest is 3.7.0.

Cheers, Peter
--
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. [phone]

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist@...
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

[Andreza Vieira]

Peter Reutemann-3 wrote:

> > Sorry, I have installed the 3.6.1 version, but I'm not finding the
> > "treatZeroAsMissing" in the Apriori algorithm parameters. Where can I find
> > it?

Hmm... Quite possible that it's only available in the developer
version (I never used any other version). The latest is 3.7.0.

Yes, it's available in the developer version 3.7.0, thanks =)
More one doubt: in this version exists the parameter "removeAllMissingCols" (Apriori algorithm). Does I also have to set it to "true"?

Thanks a lot.

Cheers, Peter
--
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. [phone]

_______________________________________________
Wekalist mailing list
Send posts to: [e-mail]
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

[Peter Reutemann-3]

> Yes, it's available in the developer version 3.7.0, thanks =)
> More one doubt: in this version exists the parameter "removeAllMissingCols"
> (Apriori algorithm). Does I also have to set it to "true"?

You can, as it removes columns that are completely useless (columns
that only contain missing values).

Cheers, Peter
--
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. [phone]

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist@...
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

[Andreza Vieira]

Peter Reutemann-3 wrote:

> Yes, it's available in the developer version 3.7.0, thanks =)
> More one doubt: in this version exists the parameter "removeAllMissingCols"
> (Apriori algorithm). Does I also have to set it to "true"?

You can, as it removes columns that are completely useless (columns
that only contain missing values).

Thanks for helping.
Now, I'm using Classification functionality in the Weka.
I want to know if there's any Classification algorithm that I can set a parameter (as we can do with Apriori: treatZeroAsMissing) to consider only the attributes with the values = "true".

Thanks a lot,
Andreza

Cheers, Peter
--
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. [phone]

_______________________________________________
Wekalist mailing list
Send posts to: [e-mail]
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

[Peter Reutemann-3]

>>> Yes, it's available in the developer version 3.7.0, thanks =)
>>> More one doubt: in this version exists the parameter
>>> "removeAllMissingCols"
>>> (Apriori algorithm). Does I also have to set it to "true"?
>>
>> You can, as it removes columns that are completely useless (columns
>> that only contain missing values).
>>
>
> Thanks for helping.
> Now, I'm using Classification functionality in the Weka.
> I want to know if there's any Classification algorithm that I can set a
> parameter (as we can do with Apriori: treatZeroAsMissing) to consider only
> the attributes with the values = "true".

Don't think so that there is any classifier with that highly specific
functionality (typical only for associators, not classifiers).

Cheers, Peter
--
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. [phone]

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist@...
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html