Is giving directions to an Agent trough Machine Learning
What can be the motivation for using ML in the execution phase?
What technics can I apply?
Depends on the required output.
Is it technically difficult?
Yes! as Agent-based modelling and machine learning needs to be fully integrated.
After the first initialization of the ABM model, agents start to collect observations and send them to the ML algorithm.
The algorithm recommends specific actions and records them in its history.
This recommendation is sent back to the agent, adjusting its behavior and creating a new observation.
The cycle starts again.
(Rand W., 2006)
It requires data that can be compared and stored in a single record comparable in both the ABM model and ML Algorithm.
The output of the ML algorithm is an action that can be transformed into a function for the agent behavior.