Sequenced Decision Trees

Adriana_Cira09Adriana_Cira09 Member Posts: 1 Newbie
edited July 2020 in Help
Hello, 
I am trying to create a decision tree model for a dataset consisting of law cases. I have extracted the relevant information and now I want to create a decision tree that follows the following pattern: Case Categ -> People Categ -> Procedure Categ -> Result. I want the model to go through each category in this specific order. However, I have no idea how to do that in RapidMiner. Could someone help me?
I should mention that in each category, there are variables: Case (age of the offender, gender, gravity, circumstances), People (names for the judge, lawyers, prosecutor), Procedure (location, court, accusing entity). And the final result will be whether or not the person was acquitted and if it was not acquitted, how many years did it get.
Tagged:

Answers

  • Telcontar120Telcontar120 Moderator, RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 1,635 Unicorn
    There is no way to force a specific sequence of attributes via the DT algorithms in RapidMiner.  You can limit the attributes that the DT will consider, but attributes and splits are selected based on the algorithmic parameters of the DT implementation, not user selection.
    It sounds like what you really want is just a table of specific conditional probabilities.  You can replicate that easily in RapidMiner by simply using the Aggregate operator.  In your case, enter the attributes you want to consider as the grouping attributes and then the case outcomes as the aggregation attributes (you can do counts and if you code acquittal as a 0/1 variable you can use average to compute the acquittal rate, and then average of the number of years sentenced). 
    Brian T.
    Lindon Ventures 
    Data Science Consulting from Certified RapidMiner Experts
Sign In or Register to comment.