I'm working on a genetic algorithm concerning a timetabling problem, and I have the problem of deciding which selection method I should choose.
My questions are:
1)How the choice of a selection method can affect the performance of a genetic algorithm?
2)What is the most performant method of selection?
3)What happens if I try to use a random selection to evade fitness-based selection (because this task in my case takes time)?
Thank you in advance ..
1) Your fitness function, mutation/crossover operators, and overall search space are intimately and subtly interconnected. Each problem/selection-method will have it's own performance characteristics. So, there are no definitive answers to your question. This is still an open research question.
2) Can't say because of 1). You'll have to try, measure, & see for yourself.
3) Probably a complete random walk as ziggystar suggests, but that could be ok.