Contribution of the predictors to the target variable, ROC curve Editing,

SA_H

Please I have two questions:

1- How can I estimate the contribution of the predictors to the target variable in a binary classification problem.

2- How I can edit the figures exported from rapidminer and in particular ROC curve.

Find more posts tagged with

AI Studio

AUC ROC

Accepted answers

All comments

varunm1

Hello @summer_helmi

If you are looking to understand the contribution if an attribute (predictor) on the prediction, RM has a "explain predictions operator", this operator will provide you with the predictors that supported and contradicted each prediction label (might be a correct or wrong prediction). This operator will calculate the local correlation of each predictor on the predicted label.

You can use online image editors for editing images or adobe tools.

Thanks

Telcontar120

Some ML algorithms directly output global independent variable importance as well. It just depends on the algorithm you have chosen. Explain Predictions generates local estimates which are good for understanding specific cases but not necessarily for understanding which ones are the most important overall.

IngoRM

...but not necessarily for understanding which ones are the most important overall.

100% agreement. In many cases it becomes "visually" obvious though since the globally important columns "stand out" since they have bolder colors in most cases. I saw that now quite often and I am thinking about putting this into an algorithm right now...

But you can also use any of the "Weighting" operators to calculate global importance BTW.

Just my 2 cents,
Ingo

SA_H

Thank you all for your replies. I will try the suggested operators.