Ant Colony Optimization for Winner Determination in Combinatorial Auctions

Abstract

Determining the winners of combinatorial auctions which maximize the profit of the auctioneer is NP-complete problem. This paper presents an efficient approximate searching algorithm IAA for the problem. The new algorithm uses the ant colony optimization algorithm based on heuristic rules, the proposed algorithm not only give the way for identify feasible bids with a given partial solution but also avoid the unnecessary trials that will not lead to an optimal solution. We have implemented IAA with Visual C++6.0, experiment results show IAA has good performance. When the average error of IAA is less than 2%, the running time of IAA is less than half of Edo Zurel and Noam Nisan's ALPH algorithm in random and weighted random distributions. Meanwhile IAA can get excellent solution for problem with over 3000 items and 50000 bids.