Authors: Birkan Can,Gearoid Murphy, Cathal Heavey
In this work, we investigate evolutionary metamodelling of discrete-event simulation models with the buffer allocation problems. We propose a genetic programming approach in order to derive the artificial response functions of simulation models. Alternative to similar studies, we do not assume a form for the response function and perform symbolic regression analysis over simulation models of different sizes of serial production lines. We present a comparative analysis with another artificial technique, neural networks, to identify the efficiency and the performance of symbolic regression in deriving metamodels via simulation.