A BENCHMARK FOR AUTOMATED ROSTER GENERATION ALGORITHMS

This paper describes a benchmark that enables objective comparison bet ween the implementations of algorithms for automated shift roster gene ration. The benchmark consists of three computational tests that provi de measures of correctness, efficiency and efficacy. The tests are des igned to ensure that it is difficult to specifically fine tune an impl ementation to perform well on the benchmark. A recently developed impl ementation (Bohle and Harris, 1996) is used to provide run-time perfor mance metrics for a variety of PC hardware configurations.