Data Mining and Predictive Analytics, 2nd Edition

Chapter 27Genetic Algorithms

27.1 Introduction To Genetic Algorithms

Genetic algorithms (GAs) attempt to computationally mimic the processes by which natural selection operates, and apply them to solve business and research problems. Developed by John Holland in the 1960s and 1970s (Holland¹), GAs provide a framework for studying the effects of such biologically inspired factors as mate selection, reproduction, mutation, and crossover of genetic information.

In the natural world, the constraints and stresses of a particular environment force the different species (and different individuals within species) to compete to produce the fittest offspring. In the world of GAs, the fitness of various potential solutions is compared, and the fittest potential solutions evolve to produce ever more optimal solutions.

Not surprisingly, the field of GAs has borrowed heavily from genomic terminology. Each cell in our body contains the same set of chromosomes, strings of DNA that function as a blueprint for making one of us. Then, each chromosome can be partitioned into genes, which are blocks of DNA designed to encode a particular trait such as eye color. A particular instance of the gene (e.g., brown eyes) is an allele. Each gene is to be found at a particular locus on the chromosome. Recombination, or crossover, occurs during reproduction, where a new chromosome is formed by combining the characteristics of both parents' chromosomes. Mutation, the altering of a single gene in a chromosome ...