Bidding Optimization for Display Advertising under Budget Constraint

Abstract

This work is all about bidding optimization that can be used in display advertising. We study and investigate the problem of sequential second price auction that provides a partial observation to the winning agent. We also study bidding strategies can be used in a practical situation under budget constraint and uncertainty of market price for a single- sided auction. Besides, we assume the market price is random variable drawing from the gamma distribution. After that, we model and simulate the problem via MATLAB, and also decide to use the integer programming framework for the sake of simplicity of modeling, designing and evaluating the problem. Therefore, we propose our algorithms, heuristic approach, to tackle the problem that are comprised of three algorithms: Greedy, Greedy-Knapsack and Improved Greedy-Knapsack algorithms, which are improved step- by-step respectively. Therefore, we evaluate our algorithms and then compare ours with other existing algorithms such as simple, heuristic and also dynamic programming approach algorithms. The results show that the final version of greedy archive the best performance. We also evaluate the dynamic programming algorithm compared with our algorithms as a special case due to the processing time that take too long in practice. Obviously, our algorithms does not perform well as the dynamic programming does, but the results are not bad since our algorithms can be computed in polynomial time that contrast to the dynamic programming approach that computes in an exponential time.

More information

Identifiers

ORCID for Enrico Gerding: orcid.org/0000-0001-7200-552X

Catalogue record

Export record