Geoscience Reference
In-Depth Information
7
Evolutionary Algorithms
Alison Heppenstall and Kirk Harland
CONTENTS
Abstract .......................................................................................................................................... 143
7.1 Introduction .......................................................................................................................... 14 4
7.2 Brief History of EAs ............................................................................................................. 14 4
7.3 EA Family ............................................................................................................................. 145
7.4 Building Blocks of EAs ........................................................................................................ 146
7.4.1 Initial Populations ..................................................................................................... 146
7.4.2 Fitness Function ........................................................................................................ 147
7.4.3 Selection .................................................................................................................... 147
7.4.4 Recombination/Crossover ......................................................................................... 148
7.4.5 Mutation .................................................................................................................... 148
7.4.6 Single- versus Multiple-Objective Optimisation ...................................................... 149
7.5 EA Resources and Software ................................................................................................. 149
7.6 Applications of EAs in Geography ....................................................................................... 150
7.7 Example 1: Optimising Model Parameters ........................................................................... 151
7.7.1 Model ........................................................................................................................ 151
7.7.2 Rule Sets ................................................................................................................... 151
7.7.3 Parameters to Be Optimised ..................................................................................... 152
7.7.4 Optimal Solution Space ............................................................................................ 152
7.7.5 Statistical Measure of Fitness ................................................................................... 152
7.7.6 Comparison of Parameter Values ............................................................................. 153
7.8 Example 2: Breeding Model Equations ................................................................................ 157
7.8.1 Breeding SI Models for the Education Sector .......................................................... 157
7.8.2 Simplifying the SI Model Equations ........................................................................ 157
7.8.3 Representing the Equations ...................................................................................... 158
7.8.4 Calibrating the SI Model Equations ......................................................................... 159
7.8.5 Results ....................................................................................................................... 160
7.8.6 Impact of GA Calibration on the SEM ..................................................................... 160
7.9 Discussion and Conclusions .................................................................................................. 161
References ...................................................................................................................................... 163
ABSTRACT
This chapter presents
evolutionary algorithms
(EAs) with an emphasis on
genetic algorithms
(GAs) and their ability to search large areas of a solution space for finding optimal parameters.
EAs can also be used to build models by breeding model equations. We present the building
blocks of EAs followed by a review of EA applications in geography. Two illustrative case stud-
ies are then provided which demonstrate how GAs can be applied to find optimal parameters in
an agent-based model of retail and in breeding spatial interaction models for use in education.
We end with a general discussion on the utility of GAs and, more broadly, EAs in geography and
GeoComputation.
143
