Neural Nets In Technical Analysis
by Yin Lung Shih
The technical analysis of price data requires a mix of mathematical technique, experience and intuition.
Anyone with a background in high school math can understand, or at least calculate, the mathematical
functions that have been developed to transform raw price series into more meaningful charts.
Experience, on the other hand, takes time to develop, and intuition is even more difficult to obtain (and
indeed may never come at all). Yet it may be intuition, more than anything, that filters out the mass of
information available to settle on the precise combination and interpretation of technical indicators
appropriate to a trading opportunity.
What is intuition? One definition might be the unconscious application of the brain's innate
pattern-matching capabilities to shapes, symbols, quantities, events and concepts in a given situation,
resulting in the perception of previously "hidden" correlations. No conscious reasoning takes place, but a
vast amount of indirect and inferential processing does occur.
Recent work on simulated "brains," called neural networks, has opened up the possibility of supplying
this kind of learning and inference to many applications. Neural nets are emulations of biological
neurons, the most sophisticated collection of which is the human brain. While digital computers are built
around Boolean (true/false) operations, biological neurons can process a continuous range of intermediate
values (for example, not-quite, maybe, almost) as well. More important, biological neurons can learn
from experience, they detect subtle relationships between varied inputs and adapt to changing and/or