Machine Learning

Dear All, I have few queries. While applying Machine Learning algorithms to a given dataset. We use learner and predictor nodes for training and testing. My question is how to obtain the statistics performance evaluation matrices for the trained model like MAPE, R2, etc which we are obtaining for prediction or test dataset using predictor node. Also, I am not obtaining the trained data of the response variable like what we get using the predictor node. Kindly give insight into this. Is this particular thing is not available in KNIME. Like what we can get using MATLAB. Unless we do not have response variable results for both trained and tested results then its not possible to make interpretation based on only R2. Please share your views on this particular.

@Jitudeepu99 welcome to the KNIME forum. First I would like to point you to this entry about KNIME and machine learning in general, also about using Python nodes:

Then ether are the scorer nodes that would be able to give back some statistics you might need - Scorer, Numeric Scorer - this one has MAPE, H2O Numeric Scorer, Scorer Java Script and for Binary Class Problems there is the: Binary Classification Inspector.

For some models LogLoss might also be an option.

5 Likes