|M.Sc Student||Eran Dror|
|Subject||Increasing the Auctioneer's Expected Revenue in a Dynamic|
Multiple Unit Auctions
|Department||Department of Industrial Engineering and Management||Supervisor||Dr. Onn Shehory|
A multi-unit auction type called the ‘Expanding multi-unit auction’ is discussed in this paper. In an expanding multi-unit auction, one item of a good is initially offered to the bidders, the bidders are free to raise their bid whereas the auctioneer is free to raise the number of units offered by one each time a new bid is received. The auction halts when no more bids are placed. The winner list is constructed of the highest bids offered where is the number of units offered upon auction termination.
The goal of this study is to provide an auctioneer with a strategy that maximizes its expected revenues in an expanding multi-unit auction. The factor that most significantly affects the auctioneer’s revenue in an implementation of an expanding multi-unit auction is the decision in which context to raise the number of units offered and in which to preserve it as is. We analyze the auctioneer’s behavior in a given context, aiming to optimize his/her revenue. To arrive at the sought solution, we model the auction process as a state graph. The graph consists of nodes describing the states of the auction in a given context and of edges describing the transitions between the states. Once the state graph is formed, finding the optimal strategy is a matter of solving a search problem on the state graph. In our research we prove that the search problem to be solved, although seemingly exponentially bounded, is actually linearly bounded.