Skip to main content

Efficient Qualitative and Quantitative Use of Knowledge-Based Systems in Financial Management

Objective

The area of application of the planned knowledge-based systems was financial portfolio management. This is not a new area for knowledge-based computing: it was identified early in the USA as a profitable area for expert system development. However successhas been limited. Clients' needs have not been correctly appreciated, and there have been shortcomings in the standard rule-based model for knowledge representation when applied to financial problems.
EQUUS aimed to remedy this situation. Its objectives were to:
-develop an expert system for portfolio management
-provide the tools for acquiring knowledge and applying knowledge in the same area.
In consequence, a major goal was to supplement standard tools and methodologies with facilities for mixed quantitative and qualitative reasoning and to organise the collected knowledge accordingly.
The area of application of the planned knowledge based systems was financial portfolio management. The objectives were to develop an expert system for portfolio management, and provide the tools for acquiring knowledge and applying knowledge in the same area. The functional requirements for such financial system development were stated, and a generalized methodology for knowledge acquisition in the financial domain established. A survey of the French, American and British tools and applications was released. The requirement to reason qualitatively from time dependent quantitative data was approached through a study of the fundamental issues involved in applying fuzzy logic. This has been sufficiently advanced to incorporate the results in the demonstrators. A generalized methodology for knowledge acquisition in the financial domain was established. Ten prototypes allowing portfolio design exist, and are part of a general conceptual frame. The 2 main protypes are firstly, IPDS, written in OPS5: this is a portfolio design system allowing maintenance and including MPT (US Modern Portfolio Theory based on utility theory), characterizing the investment risks and return to be taken. Secondly, FOAL2, written in QSL/Lisp: this is a prototype based on fuzzy sets and necessity/possibility theory. The system is written in Vax Lisp and QSL language and runs under VAX VMS. The role of FOAL2 is to build a portfolio by sorting and selecting stocks on the basis of a client interview and by using fuzzy logic.
The functional requirements for such financial system development were stated, and a generalised methodology for knowledge acquisition in the financial domain established. A survey of the French, American and British tools and applications was released.The requirement to reason qualitatively from time-dependent quantitative data was approached through a study of the fundamental issues involved in applying fuzzy logic. This has been sufficiently advanced to incorporate the results in the demonstrators.A generalised methodology for knowledge acquisition in the financial domain was established.
Ten prototypes allowing portfolio design exist, and are part of a general conceptual frame, QUIDS (Equus Integration Decision System). The two main prototypes are:
-IPDS at Citymax, written in OPS5: this is a portfolio design system allowing maintenance and including MPT (US Modern Portfolio Theory based on utility theory), characterising the investment risks and return to be taken.
-FOAL2 at DATAID written in QSL/LISP: this is a prototype based on fuzzy sets and necessity/possibility theory. The system is written in Vax Lisp and QSL language and runs under VAX VMS. The role of FOAL2 is to build a portfolio by sorting and selectingstocks on the basis of a client interview and by using fuzzy logic.
The project has also produced documents on expert systems, evaluation and metrication. These are relevant to other evaluation projects (numbers 857 (GRADIENT), 1257 (MUSE), 1570 (ESCA) and 2148 (VALID)).
Exploitation
One of the prototypes, the chartist advisor of RIADA, has been integrated with the main database of RIADA and produces daily recommendations. Other prototypes (Citymax and DATAID) are planned to be progressively exploited in the future.
At the same time, a user handbook of procedures for the production of such systems has been produced. This includes measures of performance and cost estimates for the introduction of expert systems; this document will be published by UCL a methodology book on building KBS duplications.

Coordinator

Citymax Ltd
Address
6 Laurence Pountney Hill
EC4R 0BL London
United Kingdom

Participants (3)

DATAID
France
RIADA & CO
Ireland
Address
Grafton Street 28/29
Dublin 2
Birkbeck College, University of London
United Kingdom
Address
Malet Street, Bloomsbury
WC1E 7HX London