Graphical user interface design using genetic algorithms

This paper presents the automatic generation of graphical user interfaces (GUIs) using genetic algorithms (GAs), where simulated evolution is utilized to build GUIs according to simple pre-specified GUI design principles. A GA produces a population of solutions (called chromosomes) to a given problem by iteratively applying genetic operators to the members of this population to evolve better solutions. A mechanism was developed to represent simple GUIs using binary strings. The GUls included components such as button, label, list, checkbox, etc. Two represented characteristics of each component were its size and color. The GA used the fitness calculation, selection, crossover, combination, and mutation to generate populations of chromosomal strings representing the GUIs. Every time a new offspring was produced, GA evaluated the fitness of the solution based on an encoded form. Fitness value was computed based on predefined rules derived from GUI design principles. The initial generation was generated using random strings, and the simulated evolution was performed for 1000 generation with each generation having the population size of 250. The results showed that using genetic algorithm, GUIs were improved to be close to what would be expected from the pre-specified rules on component characteristics.