Speaker
Description
JMP software converts data into insights with no coding required, and is a leading solution for real-world problem-solving in many industries. Some users call the JMP Profiler, the key tool for any data modeler, “the coolest thing in JMP”. This presentation will demonstrate several Profilers and various use cases and discuss it’s value in both industrial settings as well as in teaching and learning.
Profilers are interactive visualizations of any model built in JMP, being tree-based, regression models, neural networks or other predictive models. The profiles are cross-section views of the response surface for any number of factors (Xs) and responses (Ys). All factors can be changed interactively to see the effects on the response(s) and on other profiles. Additional Profiler features help with the model understanding and interpretation, like confidence intervals, overlaid data points or interaction traces, sensitivity indicators and extrapolation warnings.
Based on the desiriability representing the goals for each response, like maximize, minimize or match target, the Profiler can also find the best factor settings to optimize the response(s) for the system or process at hand. A built-in Monte-Carlo Simulation and Gaussian Process model helps to find more robust settings in the light of any stochastic variation of the factors.
Beside the Predition Profiler, we will also demonstrate a Contour Profiler, an Interaction Profiler and a Design Space Profiler – all interactive and visual tools to get the most out of your models.