Search code examples
machine-learningweka

Weka - semi supervised learning - how to label data and get back the result?

I have just started to use Weka and I would like to ask you. I have installed a collective classification package and i have a simple training data.

X Y Label
--------------------
1 2 Class 1
3 2 Class 1
3 3 Unknown class
4 2 Unknown class
11 12 Unknown class
15 20 Unknown class

Is it possible somehow get data back from Weka labeled? Maybe I don't understand the semi-supervised method because in my opinion it's used to label other data if I have labeled a little subset.

In my case, I would like to annotate several normal instances, get label other similar instances and in the end detect anomaly instances..

Thank you your advices

Solution

  • My understanding is that you would like to store the predicted labels of your model into your missing labels.

    What you could do is right-click on the Model after training, then select 'Visualize Classifier Errors'. In this visualization screen, set Y as the predicted class and then save the new ARFF. This datafile should then contain the predicted and class labels.

    From there, you could try to replace the Missing Values with the predicted labels.

    I hope this assists in the problem that you are experiencing.