"Decision tree with one node in spite of low confidence and min gain"

olgakulesza2

Helo,

I have a problem with my decision tree. It generated only one node. Then I started to minimize the confidence even to 0.1 and min gain to 0.001. However, it didn't help. Could you please tell me what to do?

<宏/ >

< =“tru运营商激活e" class="process" compatibility="8.1.003" expanded="true" name="Process">

< =“tru运营商激活e" class="retrieve" compatibility="8.1.003" expanded="true" height="68" name="Retrieve Books_Ratings_Tags_forUser10" width="90" x="112" y="85">


< =“tru运营商激活e" class="split_data" compatibility="8.1.003" expanded="true" height="103" name="Split Data" width="90" x="246" y="85">





< =“tru运营商激活e" class="concurrency:parallel_decision_tree" compatibility="8.1.003" expanded="true" height="103" name="Decision Tree" width="90" x="447" y="34">



< =“tru运营商激活e" class="apply_model" compatibility="8.1.003" expanded="true" height="82" name="Apply Model" width="90" x="581" y="136">

Best wishes

Olga

rfuentealba

Hello@olgakulesza2,

I loaded your example, but don't have your data. However, I noticed that you have selected both "apply pruning" and "apply prepruning" on the parameters. You might want to adjust these settings, as these effectively reduce the amount of leaves generated in the tree.

What helps me adjusting a tree with "some" brute force: count how many columns are on the dataset and adjust the maximal depth to the amount of columns + 1. If this does not satisfy your needs, begin playing with the prepruning parameters before pruning right away. Do it adjusting the amount of leaves and divisions, and rerunning the model until you are OK with your results. A piece of advice on top of this is that you might find that Cross-Validation and Optimize Parameters used together can help creating a tree that is good enough for your data.

All the best,

Rodrigo.

olgakulesza2

Thanks for help@rfuentealba! It was about to high minimal leaf size.

Thank you!

Olga