Framework of building a decision tree with sequential / temporal features

Viewed 53

I am trying to build a user retention decision tree model.

My features include user events data on different days, for simplicity, say user events on Day 1 and Day 2.

The tree I want to have is that the upper node should be on features happen before the lower nodes. Say, the root is Day 1 events, and the branches below root are Day 2 events.

After some research, I don't think existing packages have the flexibility to constrain scope of features on different levels.

What would be the best framework or approach to customize existing packages so I can achieve the objective?

How to manually select the features of the decision tree this question is about the same thing I am hoping to do, but the answer is irrelevant.

Thanks!

0 Answers
Related