RM 9.1 feedback : Auto-Model limitation

Best Answers

• Options

  MartinLiebig Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, University Professor Posts: 3,507 RM Data Scientist

  December 2018 Solution Accepted

  Hi @lionelderkrikor ,
  i guess the answer is that the new features would overfit too much? @IngoRM ?

  - Sr. Director Data Solutions, Altair RapidMiner -
  Dortmund, Germany

  1
• Options

  IngoRM Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, Community Manager, RMResearcher, Member, University Professor Posts: 1,751 RM Founder

  December 2018 Solution Accepted

  Hi,

  Yes, indeed.  Plus we changed the validation approach a bit (see some of the other threads in the community - I will post answers there soon as well) to get to more robust estimations.  This unfortunately meant that we need more data for the validation part of the models which required to increase the limit from the 50 rows to 100. 
  

  We have looked into the statistics and it seemed that less than 3% of all AM runs have been on data sets of less than 100 rows and while we are sorry that we had to increase the limit (making the life harder for those 3% of the runs) we still believe that the improvements in validation and the addition of feature engineering justified this decision.

  Again, sorry for the inconvenience & best,
  Ingo
  

  3