Genetic algorithms are a data mining technique. They are used to winnow relevant data from large data sets to produce the fittest data or, in the context of a proposed problem, the fittest solution.
For those of us who are not computer scientists or mathematicians, a genetic algorithm may best be understood as computer based calculations based on the idea that—as in evolutionary biology, and genetics—entities in a population will over time evolve through natural selection to their optimal condition. In our biological world, those traits that best serve our survival over time survive. Genetic algorithms rely on this evolutionary idea as a metaphor: In the world of data in computational systems, the fittest data survives over time. Genetic algorithms—or sets of rules--use genetic concepts of reproduction, selection, inheritance and so forth. If you begin with a large set of data, the application of genetic algorithms will eventually have them winnowed down to those that are the most "fit." Fitness will be defined in terms of the particular problem.
Genetic algorithms have been proposed, in the realm of counterterrorism, to:
- Extract the fittest nodes (or connection points) in terrorist networks, in order to analyze and act on that knowledge;
- Determine the most optimal military or other strategy to use in a particular scenario. "fitness" in this case is determined as the ability to resolve a violent conflict scenario;
- Create models of new threat scenarios by 'evolving' the most dangerous scenarios from component parts (fitness in this case means the ability to survive existing strategies for their defeat).