Assorted Boosted Trees questions
User 2568 | 3/22/2016, 10:55:28 PM
I had a couple of questions.
- Does boosted trees use XGBoost?
- How does missingvalueaction 'auto' deal with values that are 'None'
- How does it deal with categorical variables. I'm assuming str varables are treated as categorical.
- Does it deal with Ints differnently
User 91 | 3/23/2016, 7:54:29 PM
Our boosted tree version is a fork of XGboost with some added features (such as compression) which enable it to scale to larger datasets as well as additional features that help you automatically handle features that are categorical, dictionary (sparse), and list (dense).
This post (http://forum.dato.com/discussion/1691/boostedtreesclassifier-missing-value-handling) explains how we deal with missing values.
Features of type string are treated as categorical variables.
Ints are treated as numeric. You can explicitly treat them as categorical by converting them to type string.