Generating realistic data sets for combinatorial auctions

We consider the generation of realistic data sets for combinatorial auctions. This problem has been recognized as central to enhance the contribution of the computer science community to the field. We put forward the notions of structure and budget as main guidelines towards the generation of succinct and realistic input data. We describe a computational framework for the analysis of existing algorithms against realistic benchmarks, and use it in the context of two real world scenarios, i.e., real estate and railroad track auctions. The results of this analysis suggest that the obstacles to using (one round) combinatorial auctions in real world applications might be of an economic nature rather than a computational one.