"Decision Tree and Feature selection results"

ammarghammargh Member Posts: 27 Maven
edited June 2019 in Help
Using RM 5.3.008
Prediction done using Decision tree.
I have used feature selection (Optimize selection) with backward selection. The output of this process was used as an input for X-Validation block. A 90 % accuracy was achieved.
I have found that the feature selection process has removed one feature.
I have replaced the "Optimize selection" block by "select attribute" component and removed that feature. The X-Validation accuracy has dropped down to 89%.

Am I missing something? Shouldn't I get the same accuracy since all other parameters are the same?

Thank you in advanced for your reply

Answers

  • SkirzynskiSkirzynski Member Posts: 164 Maven
    Tried to reproduce what you have described, but I get the same accuracy. Could you post an example process with a small data set?
  • ammarghammargh Member Posts: 27 Maven
    I am sorry I could not get permission to provide my data.

    I will try to generate similar data artificially

    Thank you.
Sign In or Register to comment.