Decision Tree without a label... is it possible?

Eleon0r_Galisky

My task has two parts: first I use a dataset of 11 columns plus a target binomial variable (heart attack: yes or no) to choose the best model to predict a heart attack.

I was successful in comparing the models with ROC. Among decision tree, random forest, k-NN, rule induction, and naive Bayes, I figured out that Decision Tree is the best.

The next part of the task is to apply the best model to another dataset, but here the target is missing. Now I have again the 11 columns, but no column for the heart attack, as the model is supposed to predict it.

The process will not run because the Decision Tree requires a label. The same problem appears with any of the mentioned models. So, how do I solve this?

Please keep in mind that this is a school example, thus the datasets are very clean. Sadly, I am not allowed to share it. 

I have a deadline to meet, so pleeeease help me!!

Thank all of you in advance.

lionelderkrikor

Hi @Eleon0r_Galisky,

It is the classic workflow in data science : 
 - First you train a model with a labeled dataset (in your case a decision tree)
 - then you apply this trained model to an unlabeled dataset to predict for this dataset the label.

For that you have to use the Apply Model operator.

Take a look at the process in attached file to understand what I mean.

hope this helps,

Regards,

Lionel

lionelderkrikor

Hi @Eleon0r_Galisky,

It is the classic workflow in data science : 
 - First you train a model with a labeled dataset (in your case a decision tree)
 - then you apply this trained model to an unlabeled dataset to predict for this dataset the label.

For that you have to use the Apply Model operator.

Take a look at the process in attached file to understand what I mean.

hope this helps,

Regards,

Lionel

Eleon0r_Galisky

Yes, that worked!!!

Thank you so much!!

I cannot express how happy I am for the prompt and clear answer!