Making Predictions with a Trained Model

[Abhishek Majumder]

First of all, thanks Ian and other staff for making such a wonderful introductory course!

I want to get some further idea regarding how to evaluate a trained model on test data and have a predicted values column appended to the actual file.

Referring to this link:- Making predictions

I understand this can be achieved from the explorer by checking Output predictions options, doing which we get the predictions on the test set in the results pane.

My question is, what is the simplest way to get this predicted values column appended to the original test data file using the Explorer.

I reckon this can done from the KnowledgeFlow option (as given later in the above link)? But as I'm new to Weka (thus only acquainted with the Explorer) , just wanted to know if the same thing can be done from the Explorer itself.

Thanks,

Abhishek

[Ian Witten]

On 23/05/2015, at 8:24 am, Abhishek Majumder <avish...@gmail.com> wrote:

First of all, thanks Ian and other staff for making such a wonderful introductory course!

I want to get some further idea regarding how to evaluate a trained model on test data and have a predicted values column appended to the actual file.

Referring to this link:- Making predictions

I understand this can be achieved from the explorer by checking Output predictions options, doing which we get the predictions on the test set in the results pane.

My question is, what is the simplest way to get this predicted values column appended to the original test data file using the Explorer.

Putting “first-last” in the “Output additional attributes” box (after the Output predictions options you mention) will output the attribute values of each instance. However, the format is not a pretty one.

This feature is greatly improved in the next version of Weka, Weka 3.7 (you can easily download this and try it).  Select “CSV” output and configure it by clicking the CSV; there you can determine which attributes to use and whether to output the probability distribution or just the prediction.

Using J48 on the weather data, selecting CSV output and the attributes "first-last” and just the prediction (not the distribution), you get this output:

inst#,actual,predicted,error,prediction,outlook,temperature,humidity,windy

1,2:no,2:no,,1,sunny,hot,high,FALSE

2,2:no,2:no,,1,sunny,hot,high,TRUE

3,1:yes,1:yes,,1,overcast,hot,high,FALSE

4,1:yes,1:yes,,1,rainy,mild,high,FALSE

5,1:yes,1:yes,,1,rainy,cool,normal,FALSE

6,2:no,2:no,,1,rainy,cool,normal,TRUE

7,1:yes,1:yes,,1,overcast,cool,normal,TRUE

8,2:no,2:no,,1,sunny,mild,high,FALSE

9,1:yes,1:yes,,1,sunny,cool,normal,FALSE

10,1:yes,1:yes,,1,rainy,mild,normal,FALSE

11,1:yes,1:yes,,1,sunny,mild,normal,TRUE

12,1:yes,1:yes,,1,overcast,mild,high,TRUE

13,1:yes,1:yes,,1,overcast,hot,normal,FALSE

14,2:no,2:no,,1,rainy,mild,high,TRUE

Here I just evaluated on the training set, but with a test set you would just get the output for the test set.

(Sorry for the late reply)

cheers

ian

I reckon this can done from the KnowledgeFlow option (as given later in the above link)? But as I'm new to Weka (thus only acquainted with the Explorer) , just wanted to know if the same thing can be done from the Explorer itself.

Thanks,

Abhishek

--
You received this message because you are subscribed to the Google Groups "WekaMOOC-general" group.
To unsubscribe from this group and stop receiving emails from it, send an email to wekamooc-gener...@googlegroups.com.
To post to this group, send email to wekamooc...@googlegroups.com.
Visit this group at http://groups.google.com/group/wekamooc-general.
To view this discussion on the web, visit https://groups.google.com/d/msgid/wekamooc-general/7552fce9-8ab5-421d-ac24-498e34139ceb%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Abhishek Majumder]

Thanks professor for your detailed guidance! Will definitely try out v3.7.

Eagerly looking forward to June 1st for the More Data Mining course... :-)

Regards

Abhishek