Adaptive agent bidding strategy based on stochastic modeling

Abstract

For a dynamic, evolving multiagent auction, we have developed an adaptive agent bidding strategy (called the p-strategy) based on stochastic modeling. The p-strategy takes into account the dynamics and resulting uncertainties of the auction process using stochastic modeling, and avoids the shortcomings of stochastic modeling by adaptively deciding when to use the model and when not to. Our experiments show that the p-strategy outperforms other candidate bidding strategies in a continuous double auction regardless of the status of the auction and the demography of the agent population.