Least Square in DT - other parameters don't affect outcome?

stephastepha Member Posts: 1 Newbie
Hi there! I'm using Decision Tree model to conduct a predictive 'regression' model. In order to predict numerical values, criterion has to be set to 'Least Square'. 

However, when I alter all the other parameters (maximal depth, minimal gain etc.) , none of them affects the results anymore. Whether the maximal depth was 10 or 100, the final RMSE remains the same! 

Wondering does anyone know why is this? 

Best regards, 
Extreme newbie
Tagged:

Answers

  • PapadPapad Computer Science Student Member Posts: 62   Unicorn
    Hi,
    Try a value <10 to see if there is any change. Maybe the depth of the tree you have can reach at maximum 10, so even if you choose a greater number there would be no change.
    I'm not sure for my answer but I hope it helps about maximal depth.
    sgenzer
  • Telcontar120Telcontar120 Moderator, RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 1,179   Unicorn
    Indeed, without looking at the dataset and the resulting tree model, it would be impossible to tell whether this constraint or any other parameter value you have currently would actually affect the tree outcome. You need to look carefully at the resulting tree first and evaluate it with respect to the parameter values you have set, and then try modifying one or more in the way that will actually force the tree to change.
    You might consider modifying the prepruning options, they typically can have a large impact on the final tree.
    Brian T.
    Lindon Ventures 
    Data Science Consulting from Certified RapidMiner Experts
    sgenzer
Sign In or Register to comment.