| Multi-Attribute Reverse Auction (MARA) is widely used in modern electronic procurement systems. MARA is an excellent way to automate negotiations with sellers and achieve the ideal goal of buying efficiently and with minimal costs. However, online MARA has not yet become popular because of the two following issues. (1) They are not trustworthy. Buyers and sellers can behave in an unpleasant and dishonest way to achieve benefit at the expense of competitors. Also, auctioneers may be untrustworthy and be in collusion with some bidders or disclose private information of bidders. (2) Eliciting the buyer's preferences and determining the winner are challenging tasks for MARA systems.;In order to address the second issue above, we study and extend two methods, the multi-attribute utility theory and the constrained CP-nets, for solving the winner determination problem in MARA. Thanks to the proposed extensions, the buyer can now express conditional preferences over the product attributes. In addition, he can submit his preferences qualitatively or quantitatively. We also propose a new method which enables the buyers to express both quantitative and qualitative preferences in one auction. In this way, we provide buyers with more flexibility in the specification of their purchasing requests.;Finally, we implemented our GUI IMASS agent based application using Java and Jadex. This application utilizes our auction mechanism and winner determination methods. In order to evaluate the performance of our proposed methods, we executed various comparative experiments and discussed their usefulness for MARA.;In this thesis, to promote trust in online reverse auctions, we design a novel trustworthy Iterative Multi-Attribute Semi Sealed (IMASS) auction mechanism. Our protocol solves the problem of cheating auctioneers and some of the fraudulent behaviors such as shielding, sniping and siphoning. |
