Negotiating agents for online agents

Patricia Anthony, and Ho, Chong Mun and Teo, Jason Tze Wi and Rayner Alfred, and Siow, Jacob Tian You Negotiating agents for online agents. (Unpublished)

[img] Text
Negotiating agents for online agents.pdf

Download (72kB)

Abstract

The proposed research is an extension of a previous project under SCF0003-ICT-2006. The previous project is concerned with the development of bidding strategy and selling strategy that can be used by both the bidders and sellers in online auction. The previous project was able to deliver several things: a. A refined and improve bidding strategy that employs three evolutionary computation techniques namely finding the optimum crossover and mutation rate, self-adaptive genetic algorithm and adaptive genetic algorithm. Using these three methods the performance of the bidding strategy in terms of success rate and payoff was much higher compared to the previous model that used a fixed crossover and mutation rate. b. A new bidding strategy by utilizing a Grey System Theory to predict the closing price of given auction. Using this technique, we can eliminate those irrelevant auctions quickly and focus on the auctions that will most likely guarantee a win. c. A seller agent that is able to generate a strategic reserve price that can be utilized by the seller when auctioning an item. The generation of the reserve price is based on several constraints namely the profit desired, the duration of the auction, the number of bidders and the number of competing sellers who are selling the same item. This project extended the study in the previous project by studying the effects of having too many intelligent bidder agents and seller agents in the marketplace. We are interested to find out what happen to the marketplace when our intelligent agents are pitted against other intelligent agents that use various bidding strategies. Our intelligent agents are also pitted against simulated human bidders that possessed varying characteristics such ask risk seeker, risk aversion and risk neutral.In this project the performance of these intelligent agents are first evaluated with different groups of standard bidders (simulated human bidders) which possess three different characteristics namely risk aversion, risk neutral and risk seeking. In the second stage, these intelligent agents are tested against heterogeneous standard bidders to investigate the status of the marketplace.Finally, in the last stage of the experiments, our intelligent agents are pitted against other intelligent agents that use two other strategies (greedy strategy and sniping). Based in the results that we have obtained, it was observed that the market economy is affected when we implement intelligent agents in the marketplace. The most obvious observation is the auction closing price decreases as more intelligent agents are present in the marketplace. This implies that sellers may not welcome intelligent agents in joining their auctions since their revenues are reduced. On the other hand, bidders would welcome the usage of these intelligent agents since these agents can help them in purchasing the goods desired with greater saving apart from saving them a lot of time and efforts when bidding in online auctions.

Item Type: Research Report
Uncontrolled Keywords: Sellers , online auction , bidder agents , seller agents , marketplace
Subjects: H Social Sciences > HF Commerce
Depositing User: Noraini
Date Deposited: 01 Aug 2019 00:30
Last Modified: 01 Aug 2019 00:30
URI: http://eprints.ums.edu.my/id/eprint/23202

Actions (login required)

View Item View Item