The Trader's Dilemma.
Advanced trading strategies require detection and recognition of non-linear, multi-variate price patterns
not easily discovered using conventional back-testing and optimization applications.
How Do Neural Networks Help?
Neural networks provide the opportunity to develop quantitatively-derived trading strategies that, unlike other programming
techniques, do not require
lucky guesswork as to the time intervals, correlations and other predictive factors to be relied
upon. Instead, the strategist may rely on his or her insights to identify various inputs that might have
predictive quality, and then allow the software to sort out their relative merit.
Ideal for Trading Applications. Neural
networks are ideally suited to analyzing the movement through time of prices, volumes, bid-ask spreads, and other reported
variables of electronically traded financial instruments. They are especially adept at detecting previously unidentified -- and even unsuspected --
patterns in the relationships among such variables. For this reason neural
nets provide a powerful quantitative tool for well-capitalized investment firms, banks and
private investors in the analysis and trading of:
● Stocks ● Futures ● Commodities ● Currencies ● Bonds ● Indices
For similar reasons, neural networks also find application in numerous other fields of research characterized by large data sets and
pattern analysis, such as data compression, virtual reality, facial and voice recognition, signal filtering and artificial life systems.
Genetic Algorithms Yield More Powerful Neural Nets.
The use of genetic algorithms enables the programmer to
breed highly sophisticated networks innately superior to those created using
more conventional computer-assisted programming techniques. Properly configured, a genetic algorithm propagates
evolutionary improvements in neural networks through succeeding generations of strategy variants. The resulting "neurogenetic"
trading tools represent a quantum advance over traditional back-testing and optimization systems. Ongoing advances
in the power and speed of low-cost computers, combined with the proliferation of affordable, readily available single-user neural net and genetic
programming packages, have finally brought this technology within reach even of the average individual trader. When implemented in combination with either proprietary
or off-the-shelf, commercially available market trading products, these tools provide a powerful addition to the trader's arsenal.