An improved Ant Colony Algorithm for Optimal Winner Determination Problem in Reverse Combinatorial Auction

Reverse Combinatorial auctions, that is, auctions where bidders can bid on the supply right of combinations of items, tend to lead to more efficient allocations than traditional auction mechanisms in multi-item auctions where the agents' valuations of the items are not additive. A common model is formulated for optimal the winner determination problem of reverse combinatorial auction. A PRIM (Preprocessing Rules-Based Improved MMAS) algorithm is adopted for the problem. Three preprocessing rules was proved to be feasibility. By applying two of the preprocessing rules before serching begins, noncompetitive bids can be removed efficiently and the bid scale was decreased. By using the other preprocessing rule in every round of serch, unnecessary trials will be avoid early. Simulation results show that the PRIM algorithm achieves good performance than MMAS in different size problems. Meanwhile PRIM performs more steadly than MMAS as the number of bidders increase. KEY WORDSReverse combinatorial auction; Online procurement; Winner determination problem; MMAS; Preprocessing rule.