Loop example for Boosting

Hi All ,

I need help in bagging n boosting as i am new to all this . can someone show me with some example screeshot may be how to construct a loop (loop start n end) and use boosting with it and also bagging example ?

Thanks

Hi Danyal,

did you already take a look at our example server? There is a category (011_FlowVarsAndLoops) which contains example workflows about looping.

We don't have a endsemble category there (but there will be one quite soon) but you can use the predefined metanodes "Bagging", "Boosting Learner"/"Boosting Predictor" and "Delegating"

If you open these metanodes you will see how they use the looping nodes for delegating.

Cheers, Iris

hI IRIS thankx for yr resposne , can you plz give me the link of that example server , Sorry i cant find it .

Access it from within knime by clicking "connect" button in  "Server workflow Projects " tab. If this tab is not visible enable it from the View menu.

thanks nsilico , i have checked that , it does have looping example but i was lookinf for specific looping example for BOOSTING LOOP STAR  , BOOSTING LOOP END and how to use them with BOOSTING LEARNER N PREDICTOR . is there any example someone did before ?

any example diagram (snap shot ) of bagging or boosting with any two classifiers wud help .

please do let me knw

Hi danyal,

 

I will atach you a simple workflow with our ensemble meta nodes. By double clicking on the meta nodes you see the included subworkflow.

 

Cheers, Iris

Wow Iris  , this is exactly the thing which i was looking for , spot On , thankss so much for this .

One more question , can we use multiple models in bagging as it is using DT only (for about 10 times ), can we iterate with 2 or more model like NB & ANN . for example i want to do bagging with 3 model i.e. DT , NB n ANN , mean it wud have 3 columns and majority vote decides the classification .  Is this possible ? 

 

Regards

Danyal

Sure you can exchange the model with another model.

I don't see the usecase of using different models in one ensemble algorithm. How would you decide for which part of the data (bagging) you use which model? Randomly?

Regards, Iris