Abstract: Spectrum Auctions are a famous example of combinatorial auctions in practice, used almost all over the world and yielding millions of dollars revenues for the auctioneers. Researchers try to improve these auctions and their research is often based on auction simulations.
To be able to run simulations of spectrum auctions, one has to generate problem instances which consist not only of licenses (including frequency and location), but also realistic value functions of bidders. The latter is a particularly difficult task, as value functions are very complex, being influenced by many different value drivers. Additionally, there are no real world value functions we could use (the value functions are private knowledge of the bidders). In my master project, we are currently developing a Spectrum Auction Test Suite which generates problem instances based on a set of different value models.
In this seminar presentation, I will give an overview over the most important available real world bidding data and value drivers. Additionally, I will present a first version of our Java Spectrum Auction Test Suite.