Scoring - Using Classification models

mario_sarkmario_sark Member Posts: 13 Contributor I
edited June 2020 in Help

i will try to build a scoring model using Rapidminer (Classification Models). is it possible in rapidminer to have score like below for each significant variable? so the higher the score is the more the predicted value will be a YES. 

Attribute Condition  Score 
Age 18 < x<=30 6
Age >30 -2
Domiciliation Yes  5
Issurance No -2

For example in this case if the age is between 18 and 30 and the customer has a conciliation (flagged as yes) the customer score will be 11.

Thank you , 



  • Options
    kypexinkypexin Moderator, RapidMiner Certified Analyst, Member Posts: 291 Unicorn
    Hi @mario_sark

    There is no default method or operator to produce a table like this unless you assign scores manually.
    My question here is, however, what is the expected input and output of a classification model you mentioned? What kind of a dataset do you want to start with?

    Your example looks more like a traditional scoring card where each variable value (Yes, No) or bin (18-30, 30+) is assigned a specific score value, and those are summed up afterward. How do you plan though to come to certain variables binning and assigning scores?
  • Options
    mario_sarkmario_sark Member Posts: 13 Contributor I

    Thank you for your reply, actually the idea is generating scores instead of Coefficients. The aim is to target customer who do not  has deposits  and have the same behavior like those who has deposits. 

    Multiple variable we will take into consideration such as: Age, Domiciliations, Demographic,... and apply a classification technique to differentiate the two classes and identify significant variables. 

    As results customers who have scores high with no Deposits we should target them. 

    I don't know if its possible to generate these scores using RapidMiner. 

    Thank you, 
  • Options
    IngoRMIngoRM Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, Community Manager, RMResearcher, Member, University Professor Posts: 1,751 RM Founder
    Ok, I must admit that I am a bit confused now.  So let's break this down a bit :) Question 1:
    and apply a classification technique to differentiate the two classes
    What are the two classes you refer to?  In the table above, I see numerical scores, no classes so far.
    Question 2: can you please provide a simple example of a) the input you have and b) the output you would like to get.  Like one row of data and what you would like to get back for this.  Something like that :)
    I am with Vlad here that there is a high likelihood that you are actually looking for something called "Balanced Scorecard" here.  RapidMiner could be used for that, but would also be a bit of overkill.  Excel is enough for that.  But there is also a change that you actually want to train a classification model and are looking for the confidences as output.  Hard to tell at this moment, because, you know, see the first line of my response ;)
  • Options
    Telcontar120Telcontar120 Moderator, RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 1,635 Unicorn
    edited March 2019
    Actually @IngoRM I would argue that using RapidMiner is still better to generate results using balanced scorecards than Excel!  For lots of reasons relating to scalability, reproducibility, and auditability of results---e.g., all the reasons why a visual workflow and associated process are better than spreadsheets, where everything is done (generally speaking) manually or is automated in error-prone ways.  I think you've actually given presentations on this topic :smile:
    But certainly it is overkill from the perspective of machine learning, which might not be needed in this instance, depending on what the OP actually wants to do.
    Brian T.
    Lindon Ventures 
    Data Science Consulting from Certified RapidMiner Experts
  • Options
    SGolbertSGolbert RapidMiner Certified Analyst, Member Posts: 344 Unicorn

    you can easily generate said score with the Generate Attributes operator. It still remains open whether this makes sense in the context of machine learning, a learning algorithm can figure out these scores on its own based on the label, possibly better than the heuristic.


Sign In or Register to comment.