An improved evaluation method for interactive genetic algorithms and its application in product design
Yan Sun, Wang Wanliang, Xiaojian Liu · 2010
Regarding the fatigue problem in the interactive genetic algorithms (IGA) during the user's evaluation process, the paper proposed a Boolean evaluation method, which replaces user's grading work with selection. The new method reduces user's labor considerably, and makes it possible to construct user's image model (by means of user feature curves) through their selections, from which user's sensitivity to genes can also be calculated in quantified way. User image helps to estimate individual's sufficiency. User sensitivity parameters help to arrange the searching order while layered optimizing method is adopted, and can also decide which dimensions of the searching space can be neglected according to the preset threshold value. The sufficiency value's deduction method from selection information is given, together with the general skills for the recognizing and handling of coupling genes. The algorithms are tested in a product styling design example.