Brought to you by EarthWeb
IT Library Logo

Click Here!
Click Here!

Search the site:
 
EXPERT SEARCH -----
Programming Languages
Databases
Security
Web Services
Network Services
Middleware
Components
Operating Systems
User Interfaces
Groupware & Collaboration
Content Management
Productivity Applications
Hardware
Fun & Games

EarthWeb Direct EarthWeb Direct Fatbrain Auctions Support Source Answers

EarthWeb sites
Crossnodes
Datamation
Developer.com
DICE
EarthWeb.com
EarthWeb Direct
ERP Hub
Gamelan
GoCertify.com
HTMLGoodies
Intranet Journal
IT Knowledge
IT Library
JavaGoodies
JARS
JavaScripts.com
open source IT
RoadCoders
Y2K Info

Previous Table of Contents Next


2. BACKGROUND

There are several reasons why the area of finance has become such a popular a domain for expert system applications. Finance deals, to a large extent, with uncertain data, but most problems in finance can be decomposed into quantitative parts and qualitative parts and, to some extent, formulated as rules and facts, cases, or as frames, or semantic networks. The finance domain can, therefore, clearly benefit from the application of expert system technology.

Hayes-Roth and Jacobstein (1994, p .31) list the following benefits for building expert systems, all of which hold for finance applications, some examples of which are in italics:

  • Order-of magnitude increases in speed of complex task accomplishment

    XSEL reduced a 3-hour system configuration task to 15 minutes

  • Increased quality
  • Reduced errors
  • Decreased personnel required

    Canons Optex camera lens design system has made scarce highly skilled lens designers 12 times more productive

  • Reduced cost
  • Reduced training time
  • Improved decisions

    American Express claims its Authorizer's Assistant system that evaluates whether to grant or deny credit to its customers has reduced decisions to deny credit by one third, which the company estimates is worth over $27 million per year

  • Retention of volatile or portable knowledge
  • Improved customer service

    The British Government's DHSS PC-based Performance Analyst System reduced the time required for an evaluation task from 2 hours to 9 minutes -- a factor of 80 productivity gain

As a specific subdomain example, credit granting assessment systems lend themselves very well to the adoption of expert system technology, decision-making problems in the area often being too complex for handling with conventional methods. The judgment procedures are often inherently nondeterministic, and assessment also involves making judgments based on uncertain data. Moreover, the reasoning often takes place on several levels and has therefore to be decomposed. The process then leads to separate judgments for each level. For the final assessment, the subassessments have to be weighted according to some defined heuristic or rules.

3. APPLICATIONS

3.1. STOCK OPTIONS PRICING -- AESOP

The AESOP system is an attempt to apply expert system methodology to the domain of stock options pricing to be used at the American Stock Exchange. At the time when AESOP was developed, many of the options specialists at the American Stock Exchange Market (AMEX) used a classical model for valuing options developed by Black and Scholes. The Black-Scholes model arrives at a theoretical options price based on a number of assumptions (Clifford et al., 1992):

  1. A known and constant interest rate
  2. A stock price following a random walk with a variance proportional to the square of the price
  3. The stock pays no dividend
  4. The option is exercised only at expiration
  5. There are no transaction costs
  6. One can borrow to purchase or hold at the interest rate (see 1 above)
  7. There are no penalties to selling short, i.e., selling without owning the security

AESOP integrates the Black-Scholes mathematical model with a symbolic model in the form of an expert system. It provides recommendation quotations for the specialists that are closer to what they can post than the theoretical prices produced by the mathematical model alone. AESOP is an example of an expert system that has to operate close to "real-time" as the market changes.

3.1.1. The Symbolic Model

The symbolic part of AESOP represents the pricing strategy of a specialist on the AMEX, a domain that may not be receptive to mathematical modeling. The symbolic model always considers the specialist's desired spreads (the difference between bid and ask price) and always applies the specialist's rounding rules. It takes the output of the mathematical Black-Scholes model and adjusts it to incorporate the pricing strategies, with the goal of recommending bid and ask prices for each put and call for particular option series assigned to the specialist.

If the specialist's position in any series exceeds a threshold level, the symbolic model adjusts the price of that option to encourage (specialist is long) or discourage (specialist is short) trading. The symbolic model also looks for limit orders and adjusts the bid/ask prices based on the presence of these orders. Limit order adjustments are the most complicated and, potentially, the most valuable feature of the symbolic model.

The symbolic model always checks the AMEX rules to be sure that exchange regulations are not violated, and also scans for arbitrage possibilities. In almost all cases, arbitrage arises because bid/ask prices have been adjusted away from the theoretical price for some reason, most often because of the presence of a limit order.

3.1.2. Knowledge Representation

AESOP uses rules to represent the knowledge of a senior specialist at the AMEX. The developers considered this a natural approach to knowledge representation given that the American Stock Exchange has a series of rules that apply to option prices. They also observed that the heuristics used by the expert specialist seemed to follow an if-then structure, for example, "if I am long on contracts, then reduce the asking price by one increment."

3.1.3. Rules

The symbolic model of AESOP is represented using over 3000 lines of Prolog with nearly 200 rules (predicates) consisting of 400 clauses and nearly 1500 terms. Some of these predicates capture the pricing rules used by the specialist and constitute the knowledge base of AESOP, while other predicates are used to control the overall execution of AESOP, in particular the priority of various rules in the overall pricing strategy. The rules comprise Limit Order Rules, AMEX Rules, and Arbitrage Rules.


Previous Table of Contents Next

footer nav
Use of this site is subject certain Terms & Conditions.
Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details.