| Summary: | Financial markets today are facing explosive growth in the volume of market information,
global scope of risks, as well as in the addition of new mathematical complexities
embedded into financial instruments. The Thesis first discusses how analysis in financial
markets today are increasingly hindered by such operational challenges in both business
and technical aspects. Rapid financial decision-making processes have to be sustained in
the marketplace through heavier reliance on emerging computational technologies over the
next few decades.
The Thesis studies the use of neuro-fuzzy techniques as alternatives for financial forecasting.
Neuro-fuzzy computing is a hybrid technology that snaps the key strengths of both
neuro-computing and soft fuzzy-computing techniques into a single possible approach suitable
for tracking financial patterns in a real-world with great uncertainties. A major assumption
in the research work was that financial market trends exhibit time-varying characteristics.
Therefore, the Thesis develops a novel neuro-fuzzy network that embeds a
self-reorganizing learning algorithm to help reorganize fuzzy-rule structures in real-time
dynamic environments. Smart computing technologies need self-reorganizational skills to
continuously restructure fuzzy-rule structures when existing structures fail.
The Thesis recognizes the uncertainty involved in the study of real-world problems, and
does not take the conventional crystal-ball approach (which attempts to remove uncertainty)
to analyse financial market movements and trends. Fluctuations of asset prices in capital
market are the result of underlying economic behavioral patterns exhibited by investors
and speculators. This work recognizes such uncertainty in forecasting in financial technical
analysis, and the main objective of the Thesis is to improve the currency of forecasts to
support the process of financial decision making under time-varying environments.
|
|---|