Optmise deicision tree parameters

sinead_bracken

I have created a process which trains a decision tree model. 

 

The optimise parameter operator was used to optimise this model. 

 

Question: Does the optimise operator create the most satisfactory model based on the parameters you put into the operator?

i.e. if I put in min_leaf_size as min:1 and Max 15, the operator choses the most appropirate min_leaf_size?

 

 

Also, just theoretically, if my accuracy was 71.6% for my model ,what approaches would one take to improve their model? I understand each study is case specific, but are there any general rules?

 

thank you for taking the time to read this,

Sinead

sgenzer

hello @sinead_bracken - yes Optimize Parameters cycles through all the permutations selected and chooses the parameters that result in the best performance.  Note that adding additional permutations can be very resource-consuming as it's exponentially more cases to examine.  If you open the Churn Modeling template there is actually a very good example of how these pieces all connect together.

 

 

Scott

Screen Shot 2017-11-20 at 1.21.04 PM.png