Creating FCM Models from Quantitative Data with Evolutionary Algorithms

The weights of an FCM can be adjusted or entirely learned from data, which addresses limitations when experts are either unsure or unavailable. In this chapter, we show how evolutionary algorithms can perform this optimization process. Evolutionary algorithms start with a random solution and improve it by repeatedly applying operators such as mutation, crossover, and selection. The chapter defines and exemplifies these operations in Python. When there is only one candidate solution at a time, we use single-individual algorithms. In contrast, when there are several candidates, we use population-based algorithms. In this chapter, we focus on the use of population-based algorithms to optimize FCMs, which we demonstrate via two popular solutions: genetic algorithms and CMA-ES. This chapter shows readers how to apply population-based algorithms on FCMs via reusable code, while highlighting some of the key modeling choices.