Decision Support for Multiple-Unit Combinatorial Auctions Seminar 5.12.07 Riikka-Leena Leskelä (rleskela@cc.hut.fi) Supervisors: Hannele and Jyrki Wallenius
Background What has been done Computational aspects of combinatorial auctions Design of iterative (efficient) auction mechanisms – price feedback Preference elicitation What has not been done Support for bidders in constructing new bids Especially in the multiple-unit setting, which usually ignored – price and quantity feedback
Research Design of the Quantity Support Mechanism An MILP solving for profit maximizing bids (q’s and p) that team up with other bids to become winners Testing the QSM with simulations Final cost to the buyer Efficiency of the final allocation
Main Findings QSM decreases total cost to buyer and improves efficiency Shortcoming: QSM solves for one new bid at a time It won’t suggest a good bid, if complements for the bid are not in the bid stream already
Conclusion Provide ”quantity support” for bidders Next: not considered in previous literature Next: Extending QSM for groups of bidders to overcome the problems caused by considering only one incoming bid at a time Testing with human subjects
Key References Good review of combinatorial auctions: Pekeč and Rothkopf (2003), Combinatorial Auction Design, Management Science 49: 1485-1503. About the Quantity Support Mechanism: Leskelä, Teich, Wallenius H. and Wallenius J. (2007), Decision Support for Combinatorial Bundle Auctions, Decision Support Systems 43: 420- 434.