A comparison of genetic and greedy randomized algorithms for medium-to-short-term audit-staff scheduling

Based upon an empirical survey among the 200 biggest CPA firms in Germany an hierarchical modeling framework for audit-staff scheduling with three levels has been developed. For the second level, the so-called medium-to-short-term planning, a binary optimization model is introduced which is closely related to resource-constrained project scheduling. In an extensive computational study several genetic algorithms (GA) with and without domain-specific knowledge as well as a greedy randomized algorithm (RA) are investigated. Besides introducing a generalization concerning the representation of the genes of a GA which is applicable to a wide range of recent procedures it is proved that the RA is a special case of a GA. Furthermore, we observed that a GA of a modest population size with domain-specific knowledge performed best if at least a certain number of individuals is generated.