In the classification model of decision trees, we need to choose the spiting attribute. In this process normally we used to choose the highest values of GINI index as spiting attribute, but in other cases like descent attribute selection why we choose smaller value of weighted average of gini index for descent nodes of binary attributes for the construction of decision tree.
Descent methods move based on slope. If large constanta are taken it may overshoot the solution
Hi,
Which kind of cross validation method do you apply? Using cross validation will help validate the decision that was made in your decision tree.
I personally rarely use decision trees due to their tendency of over fitting the data.Having said that you could try random forest if you have access to it. That does the decision making for you based on a decent cross validation. Other than that linear discriminant or logistic regression in combination with nested cross validation should produce results that you can trust.
Feel free to have a look at the webpage below, or feel free to get in touch.
ATB
Jens
http://www.biomarker-discovery.com/data-analysis-overview
- Badges
- Science topic
- Similar topics
- Indexes