The Scientific World Journal / 2014 / Article / Tab 5

Research Article

An Efficient Fitness Function in Genetic Algorithm Classifier for Landuse Recognition on Satellite Images

Table 5

Overall accuracy of each index varying with selection ways.

Selection wayIndex
Overall accuracy (%)
DBIFCMIPASIDBFCMIAverage

Roulette wheel selection71.6 73.2 64.9 75.5 71.3
Rank selection68.3 47.7 69.1 75.1 65.0
Standard deviation2.3 18.0 3.0 0.3

String length: 8, population: 30, crossover rate: 0.8, crossover way: two-point crossover, and mutation rate: 0.003.