While the data collected can be just used for understanding the business, the final goal of any data-driven business is to optimize the business behavior by automatically making data-based and model-based decisions. We want to reduce human intervention to minimum. The following simplified diagram can be depicted as a cycle:
Figure 02-4. The predictive model life cycle
The cycle is repeated over and over for new information coming into the system. The parameters of the system may be tuned to improve the overall system performance.
While humans are still likely to be kept in the loop for most of the systems, last few years saw an emergence of systems that can manage the complete feedback loop on their own—ranging from advertisement systems to self-driving cars.
The classical formulation of this problem is the optimal control theory, which is also an optimization problem to minimize cost functional, given a set of differential equations describing the...