nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Can a Model be scored in JMP Pro 12?

jenkins_macedo

I have developed a predictive response model for the organization that I am working for. Using historical data, we are trying to predict customers that would more likely enroll in domestic programs, if they are promoted with a campaign. All the needed data of each customer segment needed to develop the model are available. Using JMP Pro 12, I developed the model using the unique capabilities that JMP provides utilizing Fit Models, Partition (DT, BF, BT) and Neural and also used the model comparison platform to compare each model and uniquely developed an average model from all, if we so decide.

I know that JMP produced a predictive score for each individual customer that we used in the model. Given that each customer has a predictive score, those results could be used sorting and creating decile of 10%, 20%, etc depending on their predictive scores.

My question is, do we need use a seperate scoring engine apart from JMP Pro, which was used to build the model to score the model? Or given that JMP already has a predictive score for each customer use in the model development process, is it fair to use those scores?

Accepted answers

KarenC

Jenkins,

In the modeling platforms are options for saving models (be sure to select the option that saves the formula) and then you will have a formula that you can use in a new data. You can either add data on to your file that you used for building the model or put the formula in the column in a new data table (column names will have to match those in the formula unless you edit the formula).

Karen

All comments

jenkins_macedo

Does anyone knowns how this works in JMP Pro? How to score a new dataset with a model developed in JMP?

"With or without an active SAS connection, JMP Pro can output SAS code to score new data quickly and easily with models built in JMP". Source: https://www.jmp.com/content/dam/jmp/documents/en/software/jmp-pro/jmp11/jmp-pro-fact-sheet.pdf

jenkins_macedo

You developed a model in JMP Pro and you are interested in using the model to score new dataset? Is that possible in JMP? I am pretty sure it is, but just can't come across the solution. I have developed various predictive models of customers who might be more likely to response to a marketing their campaign. Now, I want to be able to score each of those customers as well as new data which include similar variables (predicted and predictors). Any idea?

KarenC

Jenkins,

Karen

jenkins_macedo

Hi there Karen,

Thanks for elevating my assurance by your comment. I did exactly that because I know that the model that I developed had a prediction formula, which was saved in the data table. To score other similar data with that same sets of variables stored in the prediction formula (which can be added to or subtracted from based on our preferences). Initially did this, but was't sure. But your comments proved and demonstrated further that that step was right and I wasn't doing the wrong think. I like the option to use the prediction formula with the same sets of variables and data type in a totally new data table other than the one used to develop the model. The reason being is that the model had a total of 201K observations and the new data might have few and more rows. Even thought we can still delete the extract row, but care needs to be taken not to screw things up (human error). Thus, copy and pasting the prediction formula from the model into a new column in the new data table with a total new sets of data with the same variables seems more perfect.

Thanks once more!!

Once recommendation for JMP Developers is to have a JMP Add-in or a section under the Analyze Tab with someone can score new sets of data using the prediction formula developed by using any of the models platforms something very similar to SPSS Score Wizard would be perfect. This is just a thought.