Building Better Predictive Models Part 2: Implementing Models

Presenter: Don McCormack


Implementing Predictive Models Using JMP and JMP Pro

See how to:

  • Handle unbalanced data 
  • Aggregate categorical data to capture characteristics relevant to modeling the problem for situations where you have large number of different values (e.g., zip codes)
  • Plan the modeling approach using domain expertise to relate the data to the response of interest
  • For categorical data, use random sampling to stratify data and use JMP Profit Matrix to assign costs to undesirable outcomes and profits to desirable outcomes
  • Build models using Recursive Partition and Bootstrap Forest
  • Publish formulas to Formula Depot
  • Tune models to examine impact of top performing variables

Resources 

  • KDD Cup 1998 Data used in the demo. From the Second International Knowledge Discovery and Data Mining Tools Competition. The competition task was to estimate the return from a direct mailing in order to maximize donation profits. 
  • JMP files used in the demo

Back to Top