Minitab Blog

Zoltar…or Minitab Model Ops? You Decide.

Written by Jim Oskins | Jul 2, 2024 7:29:09 PM

Do you remember the movie “Big” starring Tom Hanks and Zoltar, the fortune telling machine? Every generation gets a great show where an oracle predicts the future (usually solving some big problem): "Big", "The Matrix", "Money Ball", "Game of Thrones"…but none are as reliable as Minitab Model Ops.

In your work I’m sure you have countless problems that you’d love some machine learning to assist with. Many customers I work with use the Minitab Predictive Analytics Module for this very thing. A customer using predictive analytics or regression analysis can deploy their models with Minitab Model Ops. You had to shoot a quarter into Zoltar’s animatronic mouth to get him to speak. Predictive analytics is much easier to use.

The eyes of the oracle are the most valued thing in “The Matrix” (except for maybe Déjà-Vu., the black cat capable of rewriting problems and clearing everyone’s memory except for the faintest feeling of… déjà-vu!). Do you remember Agent Smith and the Merovingian searching for her while Seraph, Neo, Trinity & Co. protected her? Her model was better than others. Minitab’s TreeNet (or maybe MARS) is my favorite oracle these days at work!

“Money Ball” uses statistical models to prescribe how to roster a baseball team. This is very much like a customer asked me the other day: “I am studying multiple linear and binary logistics regression, CART, TreeNet, Random Forest, and MARS. Which methods fall into predictive and prescriptive analytics?”

I thought this was very insightful… at least as interesting as Brad Pitt & Jonah Hill rebuilding the Oakland A’s based on data. What good is a prediction if you can’t use it? Wouldn’t it be nice to prescribe how to run your product or process rather than just to know how it works? Prediction alone is only one step on the hierarchy of problem-solving. If you can hook a model up to live data and use that to control a process or prevent errors… That’s power. That’s Minitab Model Ops.

See How Minitab Model Ops Can Work For You in This Webinar:

 

Real Life Examples of Predictive Analytics in Action

Here are some examples. My wife always tells me to tell readers how to do something, don’t just tell a story in these blogs! She’s a MBB too. She gets you.

In my first job out of college we had an assembly process where a measurement device was placed in an automobile component. The installation and sensitivity depended on several things like temperature and humidity in the plant. A regression model was made to relate live environmental conditions to the risk of the component performing poorly. It’s different than control charting temp and humidity... this is a model taking both as inputs to predict the future risk as the component is still being made. It was a huge undertaking then, but now that could easily be modeled, and much more accurately, with our solutions.

A Minitab customer used predictive analytics recently to model the complex relationships of land features and farming practices to predict and prevent water contamination. We’re working on an online portal so people or companies could enter their details and predict how likely contamination would be.

It’s so easy to use predictive analytics; easier than regression (a classical tool that many of my engineer friends are familiar with) & regression lets me down often.… some of us have made y=f(x) equations since childhood (or at least let our classic TI86 calculator tell us the equations & make our graphs – remember those?).  Predictive analytics finds signals that regression cannot.

  • Just open Minitab, get your historical or observational data in there… it’s usually best if you have 100 rows or more (machine learning options partition your data thanks to model validation methods… making it tough to model with too little data).
  • Then go to Stat > Predictive Analytics (which is right below most people’s favorite 6 items, just past quality tools, just past reliability… you will find 2 machine learning algorithms in this submenu). If you have a numerical response (*your column for the big Y you are analyzing is just numbers), try CART regression. OR if you have a categorical response (like pass/fail… or scratch/dent/no damage… or severity rankings, something other than just a continuous number) then try CART classification.

 

There are more amazing algorithms than these, but these are what you have for free, installed right now (if you have Minitab version 19.2020 or later!)

Minitab Predictive Analytics and Model Ops enable the concepts presented in this blog. Please let us know if we can show you how. It would be my pleasure to walk you through more advanced algorithms than CART (which can find even more hidden signals in your data). There’s an easy button in there called Automated Machine Learning (AML). As long as you keep some critical thinking of your own, AML can just run all the models, including the classical ones, and tell you which fits your data the best. We could discuss hooking these models up to your live data too with Minitab Model Ops. You could even use the regression equation from a DOE if you were so inclined… I think this would be a useful bridge that most of my old friends from industry would consider… a way to start thinking about growing beyond the traditions we already know and love.

Want More Hands-On Experience With Predictive Analytics?