Hands-On Industrial Internet of Things

Data-driven versus physics-based model

Theoretically, it is possible to identify model parameters when the amount of data entries is equal to the number of unknown model parameters. Unfortunately, due to noise in the data and the insensitivity of the context, a greater amount of data entries is required. In any case, the required amount of data samples in physics-based algorithms is fewer than that of data-driven methods.

The following decision table will help you choose between a physics-based, data-driven, or hybrid model:

	Low reliability	High reliability
A small amount of data	No model-based analytics (simple rules)	Physics-based
A large amount of data	Data-driven	Physics or data-driven

When the ...