presented by: alexander sverdov
DESCRIPTION
Review of: An Agent Oriented Business Model for E-Commerse based on the NYSE Specialist system, paper by Kenneth Griggs. Presented by: Alexander Sverdov. The Paper Outline. Description of different auctions. Description of NYSE Specialist role. - PowerPoint PPT PresentationTRANSCRIPT
Review of: An Agent Oriented Business Model for E-Commerse
based on the NYSE Specialist system, paper by Kenneth Griggs
Presented by: Alexander Sverdov
The Paper Outline
Description of different auctions. Description of NYSE Specialist role. Proposal of an agent architecture that
serves the same purpose as NYSE Specialists.
English Auction
Reserve Price (or bid floor) below which no bids are accepted.
Auctioneer guides the bidding process upwards until a final highest bid is accepted.
Dutch Auction
Reverse of an English Auction. Auctioneer states a high starting price. The Auctioneer proceeds to reduce the
price by a fixed amount until a bid is made. The first bid is taken.
Continuous Double Auction
Multiple buyers and sellers compete in the market.
Bids and Asks are cleared continuously.
Specialist in CDA at NYSE
Specialists are employees of independent specialist firms.
Manage trading of a certain stock. Governed by stock exchange rules.
Illustration
Specialists
Play the role of Market Makers. Always willing to buy and sell at a “fair” price
(even if they don’t own shares). Ensure orderly market; control the
rising/lowering of stock prices---by possibly buying/selling against the market trends (by using their own capital to cushion sharp price changes).
Can act as regular traders.
Roles of Specialists
Agents: buy/sell shares on customer’s request (act as floor broker).
Auctioneer: Provide a market for a security. Always be ready to buy/sell.
Catalyst: The specialist keeps track of all known interest in the stock, and alerts interested parties.
Principal: Buy and sell stock for their own account (with some rules).
Specialist Book
Maintain all transactions. All market and limit orders. Provides a unique view of the market.
Intelligent Software Agents
Autonomy: agents can operate without direct intervention by humans or others.
Social ability: agents can interact with other agents and/or humans.
Reactivity: agents perceive their environment and respond in a timely fashion to changes that occur in it.
Intelligent Software Agents, Cont.
Pro-activeness: agents can exhibit goal-directed behavior by taking the initiative.
Mobility: agents can move to other environments.
Temporal continuity: agents are continuously running processes.
Proposed Architecture
Architecture Details
More Detailed Details
Four Agent Types
Trading Agent Principal Agent Notification Agent Representation Agent
Trading Agent
Invokes trading rules Matches orders Maintains Bid/Ask spread Records transactions to the tape Updates Inventory Notifies other agents when required Records direct buyer-seller trades
Principal Agent
Performs an analysis using data from trade repository.
Invokes principal behavior rules. Requests trades from the trading agent.
Notification Agent
Catalyst function. Notifies possible buyers and sellers of market
conditions Maintains and updates the notification DB and
the tape. Requests trades from the trading agent. Notifies the representation agent when
applicable.
Representation Agent
Interacts with represented buyers and sellers.
Requests trades from trading agent. Negotiates commissions Updates trade repository
Database, Knowledge, etc.
The system maintains information about all participants, and all the relevant stock data.
Trading rules are handled using an expert system.
Principal behavior is determined by neural network or some heuristic system.
Things to be addressed
Consistent agent & market semantics. A deeper understanding of specialist knowledge
and functions (process model) Agent development tools. A high level agent scripting language. An analysis of knowledge representation
techniques to be used by the agents (rule-based, Expert System shell, neural net, genetic algorithm, etc.)
Conclusion
The paper provides a first attempt at modeling the NYSE specialist role using an agent based system.
The paper does not describe an implementation, but rather a possible design for such a system.
The End.