Analysing a Hybrid Model-Based Evolutionary Algorithm for a Hard Grouping Problem
Sprache des Titels:
Englisch
Original Buchtitel:
Lecture Notes in Computer Science
Original Kurzfassung:
We present a new hybrid model-based algorithm called Memetic
Path Relinking (MemPR). MemPR incorporates ideas of memetic,
evolutionary, model-based algorithms and path relinking. It uses different
operators that compete to fill a small population of high quality
solutions. We present a new hard grouping problem derived from a real
world transport lot building problem. In order to better understand the
algorithm as well as the problem we analyse the impact of the different
operators on solution quality and which operators perform best at which
stage of optimisation. Finally we compare MemPR to other state-of-the-art
algorithms and find that MemPR outperforms them on real-world
problem instances