The overall objective of this project is to develop
a software system that will combine technical and fundamental
analysis of financial data in order to efficiently perform forecasts
and trading decision support. This will provide users with an
adequate toolkit for building sophisticated automated trading
systems of a new quality, that will reduce the cognitive overload
of the trader, by supplying him with the relevant interdependencies
and offer a wider information basis.
By using text filtering techniques and following
a predefined categorisation of events, according to their significance
for the financial markets, a mechanism will be developed that
will extract these events from news, provided on-line from a data
vendor. By applying rule-based reasoning, capturing the
know-how of an expert concerning the implication of events on
the financial markets, implied events will be deduced. In addition,
a toolkit for technical analysis will be implemented, using standard
techniques as well as neural net methods on live quotes. Finally,
a combining module will be implemented using neural network and
statistical methods, capturing the mutual interdependency of extracted
news and technical analysis output. This can be augmented with
a rule-based device for modeling heuristics leading to more reliable
trading strategies. The three decision support modules will be
integrated together with facilities for evaluation and performance
measurement.
Each of the system's modules will be equipped with
a set of market independent (generic) financial model prototypes.
These model prototypes will be formed by assembling building blocks
for the development of automated trading systems and give the
software the necessary flexibility by allowing users to select
arbitrary combinations and to create their own models and applications.
In parallel, two applications of different complexity
will be implemented to prove the adequacy and flexibility of both
software and financial models to build trading systems for various
markets. The applications built during the project will serve
as reference examples on how to develop such a trading system.
This will not only demonstrate the usefulness of the software
but will be important as a support of commercial exploitation.
The application areas chosen, trading and forecasting of interest
rates (e.g. Bond futures) and currency trading (e.g. US$/DM) have
been identified as typical areas of interest for possible clients
for the system or services.