Click here for full text:
Efficiently Generating k-Best Solutions for Procurement Auctions
Byde, Andrew; Kelly, Terence
HPL-2006-145
Keyword(s): auctions; combinatorial optimization; decision support; e-commerce; graphs; algorithms; procurement
Abstract: This paper considers the problem of generating k cheapest solutions to a class of procurement auction winner determination problems. A computationally efficient solution to this problem would enable a fundamentally new approach to decision support for procurement, based on "mining" the k cheapest solutions. However, previous methods do not scale in crucial problem-size parameters, e.g., the number of sellers. Our novel algorithm achieves an exponential performance improvement over previous methods, and scales polynomially in all natural measures of problem size. By efficiently computing k-cheapest solutions, our algorithm qualitatively expands the practical applicability of the data-exploration approach to procurement decision support.
6 Pages
Back to Index
| 
Printable version
