What do you do with your models after you build them in JMP?

[#1]

As we continue to improve JMP we need to know how you exploit the models you build in JMP. Today, JMP let's you save models as prediction formulas in new columns. JMP can also output SAS DATA step code.

We want to know what you do with models you build in JMP.

• Do you deploy models built with JMP or JMP Pro into production? And if so where do those models go? Which systems do you use?
• Do you use JMP models as-is or do you re-program them into other languages? If so what languages?
• If there was one thing that JMP could do to make it easier to take a model built in JMP and put it into a production system, what would that be?

[#2]

Jeff:

I use JMP to build models for controlling petrochemical processes: analyze process data fetched from a data historian (typically Honeywell PHD); build model; give equation(s) to process control engineers who then use them in code for distributed control systems (typically Honeywell). Having good models allows us to make smart control algorithms.

This is similar to what ellie.maricou is doing.

[#3]

Jeff,

I translate the equation into SQL code and add to a pre-existing extract of customers at time 0 and then look at those same customers intermittently going forward to see if they behave consistently with the model (e.g. increase/decrease an account). I guess you would call this a "tentative non-deployment."

Thanks.

Martin 

[#4]

I transform the formula to javascript for web deployment:

http://pega-analytics.co.uk/model-viewer/html/body-measurements.html

[#5]

David, the link you posted looks very similar to the profiler we added in JMP 12 (here is a blog I wrote where I describe some of the features for any users who might not be familiar with it: Using the Interactive HTML Profiler in JMP 12 ).  Can I ask what advantages your JavaScript implementation has over the built-in Interactive HTML version?  If it isn't meeting your needs, this could be an opportunity for us to find out how to improve it in the future.

[#6]

This is a very relevant post as it applies directly to my job. As I respond to this post, I am about to complete my fifth models all developed in JMP Pro 12 using variety of data types. Once the model has been developed in JMP, I used the model to score customers who have similar historical context as those customers whose data that were used to develop the model. Once I the scoring process is done to those customers and I than retrieved their responses from whatever products/services they were promoted with in the past and using the predicted scores from the scored data table plus their actual responses, I sort the predicted scores from largest to smallest and rank the customers into decile of 10.

For example, the most targeted customers would be those in the decile 1, follow by decile 2, 3, 4, etc computing other stats in JMP such as their # of Customers, Response Rate (RR%), Actual # Enrollments, Gains %, Surplus ($), $/Piece, Average $ Amount, etc. Using these stats, we make business decisions which customers to target and who we shouldn't.

We use this process to also assess how well the model is performing in predicting reality by including the actual responses and using the curvature of the decile, which monetarily should decile from 1-10. This extract assessment process is purely exclusive of the internal JMP model validation process. Once we are all good with the model performance and the curvature of the curve of the decile depth chart.

The next step in the process is to deploy the model. Right now,I just download the Customer IDs along with other essential data fields such as the Model Score, Model Decile, Model Name into a text file, we than gets loaded into the data warehouse, which then becomes available to the marketing department for campaign segmentation, campaign analysis, targeted marketing formulation. In the near future, plan is to work internally to use JMP open database connectivity platform to load model tables directly to the marketing data warehouse.@Jeff Perkinson

[#7]

To my Clients who are JMP users, I usually deliver a data table with the model fit saved as a script customized as far as possible. To non-JMP users I deliver in Excel the model fit as a grid table as it can be generated e.g. by "Output grid table" under the red button of the prediction profiler.

I would be much interested at an easy solution to make model equations available at the internet.

[#8]

I use the models in many ways, but the feature that would be most helpful is for calibration models.  If inverse regression would work for quadratic models it would save a lot of effort.

[#9]

I would like to export models to Excel, especially neural networks

[#10 · Accepted solution]

Thanks for all the input. This was useful in developing the Formula Depot in JMP Pro.

Please continue to respond here to suggest additional destinations for models exported from the Formula Depot.