Optimising SMOreg model in WEKA using a second dataset

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

Optimising SMOreg model in WEKA using a second dataset

Ronan Flynn

Hello,


I have created an SMOreg model using a training set of data. The list of attributes is large (2400 approx). I have a second set of data, with target values, that I want to use to optimise my model. The metric that I am interested in is the correlation coefficient. Using WEKA I am not fully clear on how to use the second data set in order to improve the performance of the original model. I would be grateful if someone could outline the steps that I should follow.


In addition, should I look at attribute selection before optimisation, given the large number of attributes?


Regards,


Ronan Flynn

Tá an t-eolas atá le fáil sa ríomhphost seo faoi iontaoibh agus tá sé ceaptha le haghaidh aird an fhaighteora bheartaithe/na bhfaighteoirí beartaithe amháin. Más rud é go bhfuair tú an ríomhphost seo go hearráideach, ná húsáid agus ná tarchuir é ar mhaithe le haon chuspóir, le do thoil; ina áit sin cuir ar an eolas muid láithreach agus scrios gach cóip den ríomhphost seo ó do chóra(i)s ríomhaireachta. Ach amháin sa chás gur comhaontaíodh a leithéid go sonrach ag ár n-ionadaí údaraithe, is le húdar an ríomhphoist amháin na tuairimí a chuirtear in iúl ann, agus ní léiríonn siad tuairim ná ní chuireann siad ceangal ar aon chaoi eile ar Institiúid Teicneolaíochta Bhaile Átha Luain. Déan teagmháil le [hidden email] nó cuir glao ar 090 6468000. The information contained in this email is confidential and is designated solely for the attention of the intended recipient(s). If you have received this email in error, please do not use or transmit it for any purpose but rather notify us immediately and delete all copies of this email from your computer system(s). Unless otherwise specifically agreed by our authorised representative, the views expressed in this email are those of the author only and shall not represent the view of or otherwise bind Athlone Institute of Technology. Contact [hidden email] or telephone 090 6468000.
_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
List info and subscription status: 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: Optimising SMOreg model in WEKA using a second dataset

Martin


Hello,


I have created an SMOreg model using a training set of data. The list of attributes is large (2400 approx). I have a second set of data, with target values, that I want to use to optimise my model.



I'm not sure that I undertand you correctly, but let me assume that you have two seperated datasets (training/test) and you need to build a model based on this data using "SMOreg" classifier. Right?
 

Using WEKA I am not fully clear on how to use the second data set in order to improve the performance of the original model. I would be grateful if someone could outline the steps that I should follow.


No need to do that manually, WEKA's "FilteredClassifier" method helps to do that. This can be accomplished by combining both training and test data in one file and load this file into WEKA through the "Preprocess" panel. After that, in "Classify" panel, choose "FilterdClassifier" which is a meta-learnee, left-click on it and select "SMOreg" as a classifier parameter of  FilterdClassifier. Then press OK to close all the dialog boxes and hit "Start" button.

You could optimizethe accuracy of SMOreg by using a "GridSearch" algorithm, and you need to make sure that "RBFKerne" is the kernel of SMOreg.


In addition, should I look at attribute selection before optimisation, given the large number of attributes?


Perhaps it is a good idea to do attributes selction. WEKA allows you perform both  attribute selection and classification processes using "weka.classifiers.meta.AttributeSelectedClassifier".

Let us know if you still have any issues. 

Regards,
Martin


 


Regards,


Ronan Flynn

Tá an t-eolas atá le fáil sa ríomhphost seo faoi iontaoibh agus tá sé ceaptha le haghaidh aird an fhaighteora bheartaithe/na bhfaighteoirí beartaithe amháin. Más rud é go bhfuair tú an ríomhphost seo go hearráideach, ná húsáid agus ná tarchuir é ar mhaithe le haon chuspóir, le do thoil; ina áit sin cuir ar an eolas muid láithreach agus scrios gach cóip den ríomhphost seo ó do chóra(i)s ríomhaireachta. Ach amháin sa chás gur comhaontaíodh a leithéid go sonrach ag ár n-ionadaí údaraithe, is le húdar an ríomhphoist amháin na tuairimí a chuirtear in iúl ann, agus ní léiríonn siad tuairim ná ní chuireann siad ceangal ar aon chaoi eile ar Institiúid Teicneolaíochta Bhaile Átha Luain. Déan teagmháil le [hidden email] nó cuir glao ar 090 6468000. The information contained in this email is confidential and is designated solely for the attention of the intended recipient(s). If you have received this email in error, please do not use or transmit it for any purpose but rather notify us immediately and delete all copies of this email from your computer system(s). Unless otherwise specifically agreed by our authorised representative, the views expressed in this email are those of the author only and shall not represent the view of or otherwise bind Athlone Institute of Technology. Contact [hidden email] or telephone 090 6468000.

_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
List info and subscription status: 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]
List info and subscription status: 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: Optimising SMOreg model in WEKA using a second dataset

Ronan Flynn
In reply to this post by Ronan Flynn

Martin,


Thank you for your reply. In summary I have 3 data sets


- a training set to generate an initial model

- a development/test set to further improve and optimise the model

- a verification set (that is completely independent of any aspect of model generation) to evaluate model performance.


I will try your suggestion.


Regards,


Ronan




From: Ronan Flynn
Sent: 07 June 2016 11:43
To: [hidden email]
Subject: Optimising SMOreg model in WEKA using a second dataset
 

Hello,


I have created an SMOreg model using a training set of data. The list of attributes is large (2400 approx). I have a second set of data, with target values, that I want to use to optimise my model. The metric that I am interested in is the correlation coefficient. Using WEKA I am not fully clear on how to use the second data set in order to improve the performance of the original model. I would be grateful if someone could outline the steps that I should follow.


In addition, should I look at attribute selection before optimisation, given the large number of attributes?


Regards,


Ronan Flynn

Tá an t-eolas atá le fáil sa ríomhphost seo faoi iontaoibh agus tá sé ceaptha le haghaidh aird an fhaighteora bheartaithe/na bhfaighteoirí beartaithe amháin. Más rud é go bhfuair tú an ríomhphost seo go hearráideach, ná húsáid agus ná tarchuir é ar mhaithe le haon chuspóir, le do thoil; ina áit sin cuir ar an eolas muid láithreach agus scrios gach cóip den ríomhphost seo ó do chóra(i)s ríomhaireachta. Ach amháin sa chás gur comhaontaíodh a leithéid go sonrach ag ár n-ionadaí údaraithe, is le húdar an ríomhphoist amháin na tuairimí a chuirtear in iúl ann, agus ní léiríonn siad tuairim ná ní chuireann siad ceangal ar aon chaoi eile ar Institiúid Teicneolaíochta Bhaile Átha Luain. Déan teagmháil le [hidden email] nó cuir glao ar 090 6468000. The information contained in this email is confidential and is designated solely for the attention of the intended recipient(s). If you have received this email in error, please do not use or transmit it for any purpose but rather notify us immediately and delete all copies of this email from your computer system(s). Unless otherwise specifically agreed by our authorised representative, the views expressed in this email are those of the author only and shall not represent the view of or otherwise bind Athlone Institute of Technology. Contact [hidden email] or telephone 090 6468000.
_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
List info and subscription status: 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: Optimising SMOreg model in WEKA using a second dataset

Martin



Martin,


Thank you for your reply. In summary I have 3 data sets


- a training set to generate an initial model


Don't build the model only on the training set, you need to build your model under the evaluation of cross-validation method where you have training and test set, to create a prediction model that doesn't perform cheating. Therefore, combine the  training set with the development/test (should has labels) in one file and load it into WEKA to build the model. After that, use FilteredClassifierr as I mentioned in my previous post.

- a development/test set to further improve and optimise the model


Combine it with the previous set in one file as I explained up. 

- a verification set (that is completely independent of any aspect of model generation) to evaluate model performance.

Load this set into WEKA's Classify panel through the option "Supplied test set". 

Regards,
Martin


I will try your suggestion.


Regards,


Ronan




From: Ronan Flynn
Sent: 07 June 2016 11:43
To: [hidden email]
Subject: Optimising SMOreg model in WEKA using a second dataset
 

Hello,


I have created an SMOreg model using a training set of data. The list of attributes is large (2400 approx). I have a second set of data, with target values, that I want to use to optimise my model. The metric that I am interested in is the correlation coefficient. Using WEKA I am not fully clear on how to use the second data set in order to improve the performance of the original model. I would be grateful if someone could outline the steps that I should follow.


In addition, should I look at attribute selection before optimisation, given the large number of attributes?


Regards,


Ronan Flynn

Tá an t-eolas atá le fáil sa ríomhphost seo faoi iontaoibh agus tá sé ceaptha le haghaidh aird an fhaighteora bheartaithe/na bhfaighteoirí beartaithe amháin. Más rud é go bhfuair tú an ríomhphost seo go hearráideach, ná húsáid agus ná tarchuir é ar mhaithe le haon chuspóir, le do thoil; ina áit sin cuir ar an eolas muid láithreach agus scrios gach cóip den ríomhphost seo ó do chóra(i)s ríomhaireachta. Ach amháin sa chás gur comhaontaíodh a leithéid go sonrach ag ár n-ionadaí údaraithe, is le húdar an ríomhphoist amháin na tuairimí a chuirtear in iúl ann, agus ní léiríonn siad tuairim ná ní chuireann siad ceangal ar aon chaoi eile ar Institiúid Teicneolaíochta Bhaile Átha Luain. Déan teagmháil le [hidden email] nó cuir glao ar 090 6468000. The information contained in this email is confidential and is designated solely for the attention of the intended recipient(s). If you have received this email in error, please do not use or transmit it for any purpose but rather notify us immediately and delete all copies of this email from your computer system(s). Unless otherwise specifically agreed by our authorised representative, the views expressed in this email are those of the author only and shall not represent the view of or otherwise bind Athlone Institute of Technology. Contact [hidden email] or telephone 090 6468000.

_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
List info and subscription status: 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]
List info and subscription status: 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: Optimising SMOreg model in WEKA using a second dataset

Ronan Flynn
In reply to this post by Ronan Flynn

Hello Martin,

 

I combined my training set with the development set and used FilteredClassifier as suggested. Within FilteredClassifier I selected SMOreg as classifier, but what option for filter should I choose (default option is Discretize, which cannot handle numeric class)?

 

In combining both the training and development sets into a single file, I assume that there is no distinction between the two data sets in generating the model. In the resulting cross-validation there is an equal chance of training or development data (both sets are similar in size) being used in the model generation and test. Combining the training and development sets results in a single large training set.  I had thought that a model would be generated using the training data and that the model would be further tweaked for better performance using the second set of development data. However, unless I missed something in my study, there is no option in FilterClassifier to do this?

 

I still have my third test set of data for ‘independent’ verification of the prediction model using Supplied Test Set in the WEKA Explorer Classify panel. The metric I am interested in is Correlation.

 

Regards,

 

Ronan

 

From: Ronan Flynn
Sent: 07 June 2016 18:54
To: [hidden email]
Subject: Re: Optimising SMOreg model in WEKA using a second dataset

 

Martin,

 

Thank you for your reply. In summary I have 3 data sets

 

- a training set to generate an initial model

- a development/test set to further improve and optimise the model

- a verification set (that is completely independent of any aspect of model generation) to evaluate model performance.

 

I will try your suggestion.

 

Regards,

 

Ronan

 


From: Ronan Flynn
Sent: 07 June 2016 11:43
To: [hidden email]
Subject: Optimising SMOreg model in WEKA using a second dataset

 

Hello,

 

I have created an SMOreg model using a training set of data. The list of attributes is large (2400 approx). I have a second set of data, with target values, that I want to use to optimise my model. The metric that I am interested in is the correlation coefficient. Using WEKA I am not fully clear on how to use the second data set in order to improve the performance of the original model. I would be grateful if someone could outline the steps that I should follow.

 

In addition, should I look at attribute selection before optimisation, given the large number of attributes?

 

Regards,

 

Ronan Flynn

 

Tá an t-eolas atá le fáil sa ríomhphost seo faoi iontaoibh agus tá sé ceaptha le haghaidh aird an fhaighteora bheartaithe/na bhfaighteoirí beartaithe amháin. Más rud é go bhfuair tú an ríomhphost seo go hearráideach, ná húsáid agus ná tarchuir é ar mhaithe le haon chuspóir, le do thoil; ina áit sin cuir ar an eolas muid láithreach agus scrios gach cóip den ríomhphost seo ó do chóra(i)s ríomhaireachta. Ach amháin sa chás gur comhaontaíodh a leithéid go sonrach ag ár n-ionadaí údaraithe, is le húdar an ríomhphoist amháin na tuairimí a chuirtear in iúl ann, agus ní léiríonn siad tuairim ná ní chuireann siad ceangal ar aon chaoi eile ar Institiúid Teicneolaíochta Bhaile Átha Luain. Déan teagmháil le [hidden email] nó cuir glao ar 090 6468000. The information contained in this email is confidential and is designated solely for the attention of the intended recipient(s). If you have received this email in error, please do not use or transmit it for any purpose but rather notify us immediately and delete all copies of this email from your computer system(s). Unless otherwise specifically agreed by our authorised representative, the views expressed in this email are those of the author only and shall not represent the view of or otherwise bind Athlone Institute of Technology. Contact [hidden email] or telephone 090 6468000.
_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
List info and subscription status: 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: Optimising SMOreg model in WEKA using a second dataset

Martin
Hi Ronan,

Please see below.


Hello Martin,

 

I combined my training set with the development set and used FilteredClassifier as suggested. Within FilteredClassifier I selected SMOreg as classifier, but what option for filter should I choose (default option is Discretize, which cannot handle numeric class)?

 
This is actually based on the nature of your data and no a standard method to be applied, but anyway, you can use "weka.filters.unsupervised.instance.Resample" filter.
 

 

In combining both the training and development sets into a single file, I assume that there is no distinction between the two data sets in generating the model. In the resulting cross-validation there is an equal chance of training or development data (both sets are similar in size) being used in the model generation and test. Combining the training and development sets results in a single large training set.  I had thought that a model would be generated using the training data and that the model would be further tweaked for better performance using the second set of development data. However, unless I missed something in my study, there is no option in FilterClassifier to do this?


By combining first and second dataset, you have now one datast called training set (there is no more two datasets) that has all your instances labelled in convenient way. Since this data will be loaded into WEKA, during classification process, stratified cross-validation method (tenfold by default) will guarantee building a healthy classification model, so you'll not have any sort of cheating in the resulting model.

There is another approach you could try instead of FilteredClassifier, which is "AttributeSelectedClassifier" classifier where you need only defining AttributeSelectedClassifier parameters (classifier, evaluator, and search method). This classifier doesn't have filter to configure. 

 

I still have my third test set of data for ‘independent’ verification of the prediction model using Supplied Test Set in the WEKA Explorer Classify panel. The metric I am interested in is Correlation.


After loading your test set data into "Supplied test set" option, select (from the "More options..." button) "PlainText" see the output predictions.

Regards,
Martin
 

 

Regards,

 

Ronan

 

From: Ronan Flynn
Sent: 07 June 2016 18:54
To: [hidden email]
Subject: Re: Optimising SMOreg model in WEKA using a second dataset

 

Martin,

 

Thank you for your reply. In summary I have 3 data sets

 

- a training set to generate an initial model

- a development/test set to further improve and optimise the model

- a verification set (that is completely independent of any aspect of model generation) to evaluate model performance.

 

I will try your suggestion.

 

Regards,

 

Ronan

 


From: Ronan Flynn
Sent: 07 June 2016 11:43
To: [hidden email]
Subject: Optimising SMOreg model in WEKA using a second dataset

 

Hello,

 

I have created an SMOreg model using a training set of data. The list of attributes is large (2400 approx). I have a second set of data, with target values, that I want to use to optimise my model. The metric that I am interested in is the correlation coefficient. Using WEKA I am not fully clear on how to use the second data set in order to improve the performance of the original model. I would be grateful if someone could outline the steps that I should follow.

 

In addition, should I look at attribute selection before optimisation, given the large number of attributes?

 

Regards,

 

Ronan Flynn

 

Tá an t-eolas atá le fáil sa ríomhphost seo faoi iontaoibh agus tá sé ceaptha le haghaidh aird an fhaighteora bheartaithe/na bhfaighteoirí beartaithe amháin. Más rud é go bhfuair tú an ríomhphost seo go hearráideach, ná húsáid agus ná tarchuir é ar mhaithe le haon chuspóir, le do thoil; ina áit sin cuir ar an eolas muid láithreach agus scrios gach cóip den ríomhphost seo ó do chóra(i)s ríomhaireachta. Ach amháin sa chás gur comhaontaíodh a leithéid go sonrach ag ár n-ionadaí údaraithe, is le húdar an ríomhphoist amháin na tuairimí a chuirtear in iúl ann, agus ní léiríonn siad tuairim ná ní chuireann siad ceangal ar aon chaoi eile ar Institiúid Teicneolaíochta Bhaile Átha Luain. Déan teagmháil le [hidden email] nó cuir glao ar 090 6468000. The information contained in this email is confidential and is designated solely for the attention of the intended recipient(s). If you have received this email in error, please do not use or transmit it for any purpose but rather notify us immediately and delete all copies of this email from your computer system(s). Unless otherwise specifically agreed by our authorised representative, the views expressed in this email are those of the author only and shall not represent the view of or otherwise bind Athlone Institute of Technology. Contact [hidden email] or telephone 090 6468000.

_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
List info and subscription status: 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]
List info and subscription status: 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: Optimising SMOreg model in WEKA using a second dataset

Eibe Frank-2
Administrator
In reply to this post by Ronan Flynn
AFAIK, there is currently no way to automatically perform model optimization in WEKA based on an explicit validation set. GridSearch uses cross-validation.

I actually need this functionality myself, so I’ll probably try to implement this in GridSearch fairly soon.

Cheers,
Eibe

> On 14/06/2016, at 9:49 PM, Ronan Flynn <[hidden email]> wrote:
>
> Hello Martin,
>  
> I combined my training set with the development set and used FilteredClassifier as suggested. Within FilteredClassifier I selected SMOreg as classifier, but what option for filter should I choose (default option is Discretize, which cannot handle numeric class)?
>  
> In combining both the training and development sets into a single file, I assume that there is no distinction between the two data sets in generating the model. In the resulting cross-validation there is an equal chance of training or development data (both sets are similar in size) being used in the model generation and test. Combining the training and development sets results in a single large training set.  I had thought that a model would be generated using the training data and that the model would be further tweaked for better performance using the second set of development data. However, unless I missed something in my study, there is no option in FilterClassifier to do this?
>  
> I still have my third test set of data for ‘independent’ verification of the prediction model using Supplied Test Set in the WEKA Explorer Classify panel. The metric I am interested in is Correlation.
>  
> Regards,
>  
> Ronan
>  
> From: Ronan Flynn
> Sent: 07 June 2016 18:54
> To: [hidden email]
> Subject: Re: Optimising SMOreg model in WEKA using a second dataset
>  
> Martin,
>  
> Thank you for your reply. In summary I have 3 data sets
>  
> - a training set to generate an initial model
> - a development/test set to further improve and optimise the model
> - a verification set (that is completely independent of any aspect of model generation) to evaluate model performance.
>  
> I will try your suggestion.
>  
> Regards,
>  
> Ronan
>  
>
> From: Ronan Flynn
> Sent: 07 June 2016 11:43
> To: [hidden email]
> Subject: Optimising SMOreg model in WEKA using a second dataset
>  
> Hello,
>  
> I have created an SMOreg model using a training set of data. The list of attributes is large (2400 approx). I have a second set of data, with target values, that I want to use to optimise my model. The metric that I am interested in is the correlation coefficient. Using WEKA I am not fully clear on how to use the second data set in order to improve the performance of the original model. I would be grateful if someone could outline the steps that I should follow.
>  
> In addition, should I look at attribute selection before optimisation, given the large number of attributes?
>  
> Regards,
>  
> Ronan Flynn
>  
> Tá an t-eolas atá le fáil sa ríomhphost seo faoi iontaoibh agus tá sé ceaptha le haghaidh aird an fhaighteora bheartaithe/na bhfaighteoirí beartaithe amháin. Más rud é go bhfuair tú an ríomhphost seo go hearráideach, ná húsáid agus ná tarchuir é ar mhaithe le haon chuspóir, le do thoil; ina áit sin cuir ar an eolas muid láithreach agus scrios gach cóip den ríomhphost seo ó do chóra(i)s ríomhaireachta. Ach amháin sa chás gur comhaontaíodh a leithéid go sonrach ag ár n-ionadaí údaraithe, is le húdar an ríomhphoist amháin na tuairimí a chuirtear in iúl ann, agus ní léiríonn siad tuairim ná ní chuireann siad ceangal ar aon chaoi eile ar Institiúid Teicneolaíochta Bhaile Átha Luain. Déan teagmháil le [hidden email] nó cuir glao ar 090 6468000. The information contained in this email is confidential and is designated solely for the attention of the intended recipient(s). If you have received this email in error, please do not use or transmit it for any purpose but rather notify us immediately and delete all copies of this email from your computer system(s). Unless otherwise specifically agreed by our authorised representative, the views expressed in this email are those of the author only and shall not represent the view of or otherwise bind Athlone Institute of Technology. Contact [hidden email] or telephone 090 6468000. _______________________________________________
> Wekalist mailing list
> Send posts to: [hidden email]
> List info and subscription status: 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]
List info and subscription status: 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
|

CVParameterSelection and model generation

Ronan Flynn
In reply to this post by Ronan Flynn


Hello,


I have used CVParameterSelection to optimise a parameter (C) in SMOreg model. Following the optimisation on my training set I get a correlation value of 0.5 with my test set. I save the model but when I reload it the following is displayed


SMOreg: no model built yet.


How can I obtain/build/save the model that was generated on completion of the CVParameterSelection?


I build a model using SMOreg under Functions using the optimised parameter value suggested by CVParameterSelection. The same training and test sets are used. However, when I evaluate the model on the test set the correlation value obtained is 0.37, not 0.5 as I expected based on the CVParameterSelection. Why is this the case? It looks like the model built under Functions-SMOreg is different from that produced under CVParameterSelection, even though I am using the suggested optimised parameter from CVParameterSelection.


Any insight would be appreciated.


Thanks in advance,


Ronan Flynn

Tá an t-eolas atá le fáil sa ríomhphost seo faoi iontaoibh agus tá sé ceaptha le haghaidh aird an fhaighteora bheartaithe/na bhfaighteoirí beartaithe amháin. Más rud é go bhfuair tú an ríomhphost seo go hearráideach, ná húsáid agus ná tarchuir é ar mhaithe le haon chuspóir, le do thoil; ina áit sin cuir ar an eolas muid láithreach agus scrios gach cóip den ríomhphost seo ó do chóra(i)s ríomhaireachta. Ach amháin sa chás gur comhaontaíodh a leithéid go sonrach ag ár n-ionadaí údaraithe, is le húdar an ríomhphoist amháin na tuairimí a chuirtear in iúl ann, agus ní léiríonn siad tuairim ná ní chuireann siad ceangal ar aon chaoi eile ar Institiúid Teicneolaíochta Bhaile Átha Luain. Déan teagmháil le [hidden email] nó cuir glao ar 090 6468000. The information contained in this email is confidential and is designated solely for the attention of the intended recipient(s). If you have received this email in error, please do not use or transmit it for any purpose but rather notify us immediately and delete all copies of this email from your computer system(s). Unless otherwise specifically agreed by our authorised representative, the views expressed in this email are those of the author only and shall not represent the view of or otherwise bind Athlone Institute of Technology. Contact [hidden email] or telephone 090 6468000.
_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
List info and subscription status: 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: CVParameterSelection and model generation

Eibe Frank-2
Administrator
Hi Ronan,

Thank you very much for reporting this!

The model is actually saved correctly, which you can verify from the command-line. The reason for the strange behaviour was a bug in the getOptions() method of CVParameterSelection, which, for some reason I don't understand, called the setOptions() method of the base classifier, thereby, depending on the base classifier, potentially resetting it. The Classifier panel of the Explorer calls the getOptions() method once the model has been loaded...

I have committed a new version of CVParameterSelection without that code into the WEKA SVN repository. This means the behaviour should be fixed in the next nightly snapshot of WEKA.

I have also committed another change to CVParameterSelection that ensures consistency of results even if the base classifier is sensitive to the order of the training instances or if it does not initialize itself properly in buildClassifier(). SMOreg is sensitive to the order of the training instances. This hopefully explains your second strange result.

It would be great if you could give the new version a try.

Cheers,
Eibe

> On 28 Sep 2016, at 12:17, Ronan Flynn <[hidden email]> wrote:
>
>
> Hello,
>
> I have used CVParameterSelection to optimise a parameter (C) in SMOreg model. Following the optimisation on my training set I get a correlation value of 0.5 with my test set. I save the model but when I reload it the following is displayed
>
> SMOreg: no model built yet.
>
> How can I obtain/build/save the model that was generated on completion of the CVParameterSelection?
>
> I build a model using SMOreg under Functions using the optimised parameter value suggested by CVParameterSelection. The same training and test sets are used. However, when I evaluate the model on the test set the correlation value obtained is 0.37, not 0.5 as I expected based on the CVParameterSelection. Why is this the case? It looks like the model built under Functions-SMOreg is different from that produced under CVParameterSelection, even though I am using the suggested optimised parameter from CVParameterSelection.
>
> Any insight would be appreciated.
>
> Thanks in advance,
>
> Ronan Flynn
> Tá an t-eolas atá le fáil sa ríomhphost seo faoi iontaoibh agus tá sé ceaptha le haghaidh aird an fhaighteora bheartaithe/na bhfaighteoirí beartaithe amháin. Más rud é go bhfuair tú an ríomhphost seo go hearráideach, ná húsáid agus ná tarchuir é ar mhaithe le haon chuspóir, le do thoil; ina áit sin cuir ar an eolas muid láithreach agus scrios gach cóip den ríomhphost seo ó do chóra(i)s ríomhaireachta. Ach amháin sa chás gur comhaontaíodh a leithéid go sonrach ag ár n-ionadaí údaraithe, is le húdar an ríomhphoist amháin na tuairimí a chuirtear in iúl ann, agus ní léiríonn siad tuairim ná ní chuireann siad ceangal ar aon chaoi eile ar Institiúid Teicneolaíochta Bhaile Átha Luain. Déan teagmháil le [hidden email] nó cuir glao ar 090 6468000. The information contained in this email is confidential and is designated solely for the attention of the intended recipient(s). If you have received this email in error, please do not use or transmit it for any purpose but rather notify us immediately and delete all copies of this email from your computer system(s). Unless otherwise specifically agreed by our authorised representative, the views expressed in this email are those of the author only and shall not represent the view of or otherwise bind Athlone Institute of Technology. Contact [hidden email] or telephone 090 6468000. _______________________________________________
> Wekalist mailing list
> Send posts to: [hidden email]
> List info and subscription status: 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]
List info and subscription status: 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: CVParameterSelection and model generation

Ronan Flynn
In reply to this post by Ronan Flynn

Hello Eibe,


Thanks for your reply. I downloaded the snapshot this morning as you suggested and simply replaced my existing weka.jar with the newer one from the snapshot (I think that is all I have to do or have I missed something?). I am running WEKA 3.6.


I still have the same problem with saving and reloading the model as described below. Perhaps I have missed some step that is required when updating from the snapshot?


Regards,


Ronan




From: Ronan Flynn
Sent: 28 September 2016 00:17
To: [hidden email]
Subject: CVParameterSelection and model generation
 


Hello,


I have used CVParameterSelection to optimise a parameter (C) in SMOreg model. Following the optimisation on my training set I get a correlation value of 0.5 with my test set. I save the model but when I reload it the following is displayed


SMOreg: no model built yet.


How can I obtain/build/save the model that was generated on completion of the CVParameterSelection?


I build a model using SMOreg under Functions using the optimised parameter value suggested by CVParameterSelection. The same training and test sets are used. However, when I evaluate the model on the test set the correlation value obtained is 0.37, not 0.5 as I expected based on the CVParameterSelection. Why is this the case? It looks like the model built under Functions-SMOreg is different from that produced under CVParameterSelection, even though I am using the suggested optimised parameter from CVParameterSelection.


Any insight would be appreciated.


Thanks in advance,


Ronan Flynn

Tá an t-eolas atá le fáil sa ríomhphost seo faoi iontaoibh agus tá sé ceaptha le haghaidh aird an fhaighteora bheartaithe/na bhfaighteoirí beartaithe amháin. Más rud é go bhfuair tú an ríomhphost seo go hearráideach, ná húsáid agus ná tarchuir é ar mhaithe le haon chuspóir, le do thoil; ina áit sin cuir ar an eolas muid láithreach agus scrios gach cóip den ríomhphost seo ó do chóra(i)s ríomhaireachta. Ach amháin sa chás gur comhaontaíodh a leithéid go sonrach ag ár n-ionadaí údaraithe, is le húdar an ríomhphoist amháin na tuairimí a chuirtear in iúl ann, agus ní léiríonn siad tuairim ná ní chuireann siad ceangal ar aon chaoi eile ar Institiúid Teicneolaíochta Bhaile Átha Luain. Déan teagmháil le [hidden email] nó cuir glao ar 090 6468000. The information contained in this email is confidential and is designated solely for the attention of the intended recipient(s). If you have received this email in error, please do not use or transmit it for any purpose but rather notify us immediately and delete all copies of this email from your computer system(s). Unless otherwise specifically agreed by our authorised representative, the views expressed in this email are those of the author only and shall not represent the view of or otherwise bind Athlone Institute of Technology. Contact [hidden email] or telephone 090 6468000.
_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
List info and subscription status: 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: CVParameterSelection and model generation

Eibe Frank-2
Administrator
Sorry, I should have been more precise. You will need to download the latest snapshot for WEKA 3.9:

  developer-branch.zip

Cheers,
Eibe

> On 6/10/2016, at 10:02 PM, Ronan Flynn <[hidden email]> wrote:
>
> Hello Eibe,
>
> Thanks for your reply. I downloaded the snapshot this morning as you suggested and simply replaced my existing weka.jar with the newer one from the snapshot (I think that is all I have to do or have I missed something?). I am running WEKA 3.6.
>
> I still have the same problem with saving and reloading the model as described below. Perhaps I have missed some step that is required when updating from the snapshot?
>
> Regards,
>
> Ronan
>
>
> From: Ronan Flynn
> Sent: 28 September 2016 00:17
> To: [hidden email]
> Subject: CVParameterSelection and model generation
>  
>
> Hello,
>
> I have used CVParameterSelection to optimise a parameter (C) in SMOreg model. Following the optimisation on my training set I get a correlation value of 0.5 with my test set. I save the model but when I reload it the following is displayed
>
> SMOreg: no model built yet.
>
> How can I obtain/build/save the model that was generated on completion of the CVParameterSelection?
>
> I build a model using SMOreg under Functions using the optimised parameter value suggested by CVParameterSelection. The same training and test sets are used. However, when I evaluate the model on the test set the correlation value obtained is 0.37, not 0.5 as I expected based on the CVParameterSelection. Why is this the case? It looks like the model built under Functions-SMOreg is different from that produced under CVParameterSelection, even though I am using the suggested optimised parameter from CVParameterSelection.
>
> Any insight would be appreciated.
>
> Thanks in advance,
>
> Ronan Flynn
> Tá an t-eolas atá le fáil sa ríomhphost seo faoi iontaoibh agus tá sé ceaptha le haghaidh aird an fhaighteora bheartaithe/na bhfaighteoirí beartaithe amháin. Más rud é go bhfuair tú an ríomhphost seo go hearráideach, ná húsáid agus ná tarchuir é ar mhaithe le haon chuspóir, le do thoil; ina áit sin cuir ar an eolas muid láithreach agus scrios gach cóip den ríomhphost seo ó do chóra(i)s ríomhaireachta. Ach amháin sa chás gur comhaontaíodh a leithéid go sonrach ag ár n-ionadaí údaraithe, is le húdar an ríomhphoist amháin na tuairimí a chuirtear in iúl ann, agus ní léiríonn siad tuairim ná ní chuireann siad ceangal ar aon chaoi eile ar Institiúid Teicneolaíochta Bhaile Átha Luain. Déan teagmháil le [hidden email] nó cuir glao ar 090 6468000. The information contained in this email is confidential and is designated solely for the attention of the intended recipient(s). If you have received this email in error, please do not use or transmit it for any purpose but rather notify us immediately and delete all copies of this email from your computer system(s). Unless otherwise specifically agreed by our authorised representative, the views expressed in this email are those of the author only and shall not represent the view of or otherwise bind Athlone Institute of Technology. Contact [hidden email] or telephone 090 6468000. _______________________________________________
> Wekalist mailing list
> Send posts to: [hidden email]
> List info and subscription status: 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]
List info and subscription status: 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
|

Guidance for evaluation of attributes

Ronan Flynn
In reply to this post by Ronan Flynn


Hello all,


I have a data set in which there are 120 attributes and the class is numeric. I am using a SMOreg classifier with the C value set to 9.99E-5. The attributes of the training set and the test set are normalised before training the model and subsequent evaluation. The measurement that I am interested in is Pearson's correlation coefficient.


I would like to investigate the importance or the relevance of the attributes and would be grateful for some initial guidance as to what approach I should take, given what I have outlined above. There are a number of options under Select Attributes in WEKA and some suggestions as to what approach I should use would be appreciated. Is a wrapper-based approach that takes in the classifier type the way to go? Are there other options from which I can learn more about the contribution of the attributes? Perhaps running a number of different evaluations is required in order to gain a full picture?


Regards and thanks in advance,


Ronan Flynn

Tá an t-eolas atá le fáil sa ríomhphost seo faoi iontaoibh agus tá sé ceaptha le haghaidh aird an fhaighteora bheartaithe/na bhfaighteoirí beartaithe amháin. Más rud é go bhfuair tú an ríomhphost seo go hearráideach, ná húsáid agus ná tarchuir é ar mhaithe le haon chuspóir, le do thoil; ina áit sin cuir ar an eolas muid láithreach agus scrios gach cóip den ríomhphost seo ó do chóra(i)s ríomhaireachta. Ach amháin sa chás gur comhaontaíodh a leithéid go sonrach ag ár n-ionadaí údaraithe, is le húdar an ríomhphoist amháin na tuairimí a chuirtear in iúl ann, agus ní léiríonn siad tuairim ná ní chuireann siad ceangal ar aon chaoi eile ar Institiúid Teicneolaíochta Bhaile Átha Luain. Déan teagmháil le [hidden email] nó cuir glao ar 090 6468000. The information contained in this email is confidential and is designated solely for the attention of the intended recipient(s). If you have received this email in error, please do not use or transmit it for any purpose but rather notify us immediately and delete all copies of this email from your computer system(s). Unless otherwise specifically agreed by our authorised representative, the views expressed in this email are those of the author only and shall not represent the view of or otherwise bind Athlone Institute of Technology. Contact [hidden email] or telephone 090 6468000.
_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
List info and subscription status: 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: Guidance for evaluation of attributes

Mark Hall
As you are interested specifically in SMOreg and Pearson's correlation, I'd say that the wrapper approach is one to use. It has an option to optimise correlation coefficient. The GreedyStepwise search technique has an option to generate a ranking of the attributes. This option forces the search to the far side of the search space (i.e it will keep adding attributes even when the subset "goodness" metric starts to decline). The order that attributes are added to the current best subset results in a ranking.

Cheers,
Mark.

On 27/01/17, 11:14 PM, "Ronan Flynn" <[hidden email] on behalf of [hidden email]> wrote:

   
   
    Hello all,
   
   
    I have a data set in which there are 120 attributes and the class is numeric. I am using a SMOreg classifier with the C value set to 9.99E-5. The attributes of the training set and the test set are normalised before training the model and subsequent evaluation.
     The measurement that I am interested in is Pearson's correlation coefficient.
   
   
    I would like to investigate the importance or the relevance of the attributes and would be grateful for some initial guidance as to what approach I should take, given what I have outlined above. There are a number of options under Select Attributes in WEKA
     and some suggestions as to what approach I should use would be appreciated. Is a wrapper-based approach that takes in the classifier type the way to go? Are there other options from which I can learn more about the contribution of the attributes? Perhaps running
     a number of different evaluations is required in order to gain a full picture?
   
   
    Regards and thanks in advance,
   
   
    Ronan Flynn
   
   
   
   
   
   
   
   
   
   
   
   
    Tá an t-eolas atá le fáil sa ríomhphost seo faoi iontaoibh agus tá sé ceaptha le haghaidh aird an fhaighteora bheartaithe/na bhfaighteoirí beartaithe amháin. Más rud é go bhfuair tú an ríomhphost seo go hearráideach, ná húsáid agus ná tarchuir é ar mhaithe
     le haon chuspóir, le do thoil; ina áit sin cuir ar an eolas muid láithreach agus scrios gach cóip den ríomhphost seo ó do chóra(i)s ríomhaireachta. Ach amháin sa chás gur comhaontaíodh a leithéid go sonrach ag ár n-ionadaí údaraithe, is le húdar an ríomhphoist
     amháin na tuairimí a chuirtear in iúl ann, agus ní léiríonn siad tuairim ná ní chuireann siad ceangal ar aon chaoi eile ar Institiúid Teicneolaíochta Bhaile Átha Luain. Déan teagmháil le [hidden email] nó cuir glao ar 090 6468000. The information contained
     in this email is confidential and is designated solely for the attention of the intended recipient(s). If you have received this email in error, please do not use or transmit it for any purpose but rather notify us immediately and delete all copies of this
     email from your computer system(s). Unless otherwise specifically agreed by our authorised representative, the views expressed in this email are those of the author only and shall not represent the view of or otherwise bind Athlone Institute of Technology.
     Contact [hidden email] or telephone 090 6468000.
   
   
    _______________________________________________
    Wekalist mailing list
    Send posts to: [hidden email]
    List info and subscription status: 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]
List info and subscription status: 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
|

Only get GUI Chooser when double-click on ARFF file

Ronan Flynn
In reply to this post by Ronan Flynn

Hello,


I just upgraded to Weka 3-9-1 and now when I double-click an ARFF file I only get the GUI Chooser. Previously when double-clicking an ARFF file, Explorer opened with the file loaded.


I am running Weka on a MAC with OS 10.11.6.


Regards,


Ronan Flynn



Tá an t-eolas atá le fáil sa ríomhphost seo faoi iontaoibh agus tá sé ceaptha le haghaidh aird an fhaighteora bheartaithe/na bhfaighteoirí beartaithe amháin. Más rud é go bhfuair tú an ríomhphost seo go hearráideach, ná húsáid agus ná tarchuir é ar mhaithe le haon chuspóir, le do thoil; ina áit sin cuir ar an eolas muid láithreach agus scrios gach cóip den ríomhphost seo ó do chóra(i)s ríomhaireachta. Ach amháin sa chás gur comhaontaíodh a leithéid go sonrach ag ár n-ionadaí údaraithe, is le húdar an ríomhphoist amháin na tuairimí a chuirtear in iúl ann, agus ní léiríonn siad tuairim ná ní chuireann siad ceangal ar aon chaoi eile ar Institiúid Teicneolaíochta Bhaile Átha Luain. Déan teagmháil le [hidden email] nó cuir glao ar 090 6468000. The information contained in this email is confidential and is designated solely for the attention of the intended recipient(s). If you have received this email in error, please do not use or transmit it for any purpose but rather notify us immediately and delete all copies of this email from your computer system(s). Unless otherwise specifically agreed by our authorised representative, the views expressed in this email are those of the author only and shall not represent the view of or otherwise bind Athlone Institute of Technology. Contact [hidden email] or telephone 090 6468000.
_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
List info and subscription status: 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: Only get GUI Chooser when double-click on ARFF file

Eibe Frank-2
Administrator
I might be wrong but I don't think this is supported anymore since we removed support for Apple's Java Virtual Machine.

One thing you could do is create a little dummy "app" that starts WEKA using the OS X Automator. Create a new "Application" in the Automator and add the "Run Shell Script" action to the application. Then, change the value of "Pass input:" field from "to stdin" to "as arguments". Finally, enter the following command into the text box defining the script:

java -jar /Applications/weka-3-9-1/weka.jar $1

Save the application. Now you can associate ARFF files with this new application and they should open automatically in the Explorer if you have Java installed on your system. It seems that you can only run one instance of this app at a time though...

Some information on how to associate file types with an application is here: http://osxdaily.com/2013/08/08/change-default-application-open-files-mac-os-x/

Cheers,
Eibe

> On 4 Feb 2017, at 03:30, Ronan Flynn <[hidden email]> wrote:
>
> Hello,
>
> I just upgraded to Weka 3-9-1 and now when I double-click an ARFF file I only get the GUI Chooser. Previously when double-clicking an ARFF file, Explorer opened with the file loaded.
>
> I am running Weka on a MAC with OS 10.11.6.
>
> Regards,
>
> Ronan Flynn
>
>
> Tá an t-eolas atá le fáil sa ríomhphost seo faoi iontaoibh agus tá sé ceaptha le haghaidh aird an fhaighteora bheartaithe/na bhfaighteoirí beartaithe amháin. Más rud é go bhfuair tú an ríomhphost seo go hearráideach, ná húsáid agus ná tarchuir é ar mhaithe le haon chuspóir, le do thoil; ina áit sin cuir ar an eolas muid láithreach agus scrios gach cóip den ríomhphost seo ó do chóra(i)s ríomhaireachta. Ach amháin sa chás gur comhaontaíodh a leithéid go sonrach ag ár n-ionadaí údaraithe, is le húdar an ríomhphoist amháin na tuairimí a chuirtear in iúl ann, agus ní léiríonn siad tuairim ná ní chuireann siad ceangal ar aon chaoi eile ar Institiúid Teicneolaíochta Bhaile Átha Luain. Déan teagmháil le [hidden email] nó cuir glao ar 090 6468000. The information contained in this email is confidential and is designated solely for the attention of the intended recipient(s). If you have received this email in error, please do not use or transmit it for any purpose but rather notify us immediately and delete all copies of this email from your computer system(s). Unless otherwise specifically agreed by our authorised representative, the views expressed in this email are those of the author only and shall not represent the view of or otherwise bind Athlone Institute of Technology. Contact [hidden email] or telephone 090 6468000. _______________________________________________
> Wekalist mailing list
> Send posts to: [hidden email]
> List info and subscription status: 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]
List info and subscription status: 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: Only get GUI Chooser when double-click on ARFF file

Michael Hall
> On Feb 4, 2017, at 5:29 AM, Eibe Frank <[hidden email]> wrote:
>
> I might be wrong but I don't think this is supported anymore since we removed support for Apple's Java Virtual Machine.

Apple should of contributed the code involved to the Oracle OS X java port. It should work as before for the jvm.
If it launches the application it shows that the application is receiving the ‘open document (ODOC)’ event.
The application should have a handler that is invoked to handle that file event.
If the application doesn’t get passed the event it should be a bug report.

Michael Hall




_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
List info and subscription status: 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
|

Follow up to - Only get GUI Chooser when double-click on ARFF file

Ronan Flynn
In reply to this post by Ronan Flynn

Hello Michael, Eibe,


Thanks for your replies. I also have  Weka 3.6.7 on the same machine and when I set that as the default for opening ARFF files, Explorer opens with the ARFF file loaded in Preprocess. However, with Weka 3.9 as the default, clicking on the same ARFF file only give me GUI Chooser.


I have Java 1.8 installed and also the legacy Java 1.6 from Apple, which Weka 3.6.7 requires.


It seems to me that if GUI Chooser opens when Weka 3.9 is set as the default for ARFF files that it is odd that Explorer does not open with the file loaded.


Kind regards,


Ronan Flynn




From: Ronan Flynn
Sent: 03 February 2017 14:30
To: [hidden email]
Subject: Only get GUI Chooser when double-click on ARFF file
 

Hello,


I just upgraded to Weka 3-9-1 and now when I double-click an ARFF file I only get the GUI Chooser. Previously when double-clicking an ARFF file, Explorer opened with the file loaded.


I am running Weka on a MAC with OS 10.11.6.


Regards,


Ronan Flynn



Tá an t-eolas atá le fáil sa ríomhphost seo faoi iontaoibh agus tá sé ceaptha le haghaidh aird an fhaighteora bheartaithe/na bhfaighteoirí beartaithe amháin. Más rud é go bhfuair tú an ríomhphost seo go hearráideach, ná húsáid agus ná tarchuir é ar mhaithe le haon chuspóir, le do thoil; ina áit sin cuir ar an eolas muid láithreach agus scrios gach cóip den ríomhphost seo ó do chóra(i)s ríomhaireachta. Ach amháin sa chás gur comhaontaíodh a leithéid go sonrach ag ár n-ionadaí údaraithe, is le húdar an ríomhphoist amháin na tuairimí a chuirtear in iúl ann, agus ní léiríonn siad tuairim ná ní chuireann siad ceangal ar aon chaoi eile ar Institiúid Teicneolaíochta Bhaile Átha Luain. Déan teagmháil le [hidden email] nó cuir glao ar 090 6468000. The information contained in this email is confidential and is designated solely for the attention of the intended recipient(s). If you have received this email in error, please do not use or transmit it for any purpose but rather notify us immediately and delete all copies of this email from your computer system(s). Unless otherwise specifically agreed by our authorised representative, the views expressed in this email are those of the author only and shall not represent the view of or otherwise bind Athlone Institute of Technology. Contact [hidden email] or telephone 090 6468000.
_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
List info and subscription status: 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: Follow up to - Only get GUI Chooser when double-click on ARFF file

Michael Hall
On Feb 4, 2017, at 5:00 PM, Ronan Flynn <[hidden email]> wrote:

Hello Michael, Eibe,

Thanks for your replies. I also have  Weka 3.6.7 on the same machine and when I set that as the default for opening ARFF files, Explorer opens with the ARFF file loaded in Preprocess. However, with Weka 3.9 as the default, clicking on the same ARFF file only give me GUI Chooser.

I have Java 1.8 installed and also the legacy Java 1.6 from Apple, which Weka 3.6.7 requires.

It seems to me that if GUI Chooser opens when Weka 3.9 is set as the default for ARFF files that it is odd that Explorer does not open with the file loaded.

I don’t see where Weka is prepared to handle this either for the 3.8.1 I usually currently use now or for an old 3.7.13. 
I don’t think I have a 3.6 anymore.

If you select a file and drag it to the application the application should highlight if it is prepared to accept it.
This should also require updates to the Info.plist file to work correctly.
See below for some examples from mine.
Xcode is the default application these days for updating it.
Or you can get it into XML as shown.
For correct file handling you need the plist changes and Weka would need to register the extended file handlers.
These is a possibility that if they are really slow in registering the handlers they could miss the launching events.
Another possibility for the behavior you are seeing is that at some time you have selected a arff file and chosen open with Weka as the associated application.  
That might get the application to launch from then on without Weka doing anything. But as you say the files themselves wouldn’t actually be handled.
But my understanding  is you need both the file extension entries in the plist and the registered handlers, both Apple and Oracle jvms. 

<key>CFBundleDocumentTypes</key>
<array>
<dict>
<key>LSHandlerRank</key>
<string>Default</string>
<key>CFBundleTypeName</key>
<string>Java property files</string>
<key>CFBundleTypeExtensions</key>
<array>
<string>properties</string>
</array>
<key>CFBundleTypeRole</key>
<string>Shell</string>
</dict>
<dict>
<key>CFBundleTypeName</key>
<string>Java Jar</string>
<key>CFBundleTypeExtensions</key>
<array>
<string>jar</string>
</array>
<key>CFBundleTypeRole</key>
<string>Shell</string>
<key>LSHandlerRank</key>
<string>Default</string>
</dict>

Michael Hall





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