Hybrid Modelling and Other Practical Strategies for Modelling Real Data (2026-EU-PO-2863)

This poster highlights several practical strategies adopted over the years to make models more applicable to real systems.

For a model to be useful, it must be representative of a system. And for the model to be widely adopted, users need to trust the model.

Even though JMP has a lot of great platforms and techniques to model data, the model obtained is not always fully representative of the reality and/or well-received by the potential users. Reality is complex and when modelling real data, variables can often only take positive or strictly positive values (e.g., a person’s weight and height), or they are percentages constrained between 0 and 100% (e.g. water content in a product).

Even if the model obtained seems to capture the expected trends by the subject expert matter, a model that, in some conditions, predicts a negative value for an output that is exclusively positive in the real world will quickly be disregarded. Several strategies, such as transformations and distributions and several types of hybrid modelling, have been successfully utilised to obtain more “realistic models” and will be explained in this poster.