A comparative morphologic analysis of benchmark sets of project networks

The performance of methods to manage projects depends heavily on the features of their project networks. This is particularly true for methods devoted to project scheduling, risk analysis and resources allocation. Therefore, a long line of research has been developed to generate benchmark sets of project networks and several sets have been proposed in the literature. Unfortunately, no comparative analyses of their features were published and hence serious doubts about the comparability of results using different benchmark sets can be raised. In this paper, a multi-dimensional taxonomy for the morphology of project networks is used and four benchmark sets are evaluated: Patterson collection of problems and the sets produced by the generators due to Agrawal et al., Kolisch et al., Tavares, and Tavares et al. Original results about the lack of representativeness of these sets are obtained showing that misleading conclusions can be deduced. The last set is, by far, that one covering most extensively the morphologic space of instances which could be foreseen because the generation of networks is carried out in terms of a wider range of parameters. This conclusion is quite useful for project managers willing to assess alternative methods to solve their problems based on project networks.