When user defines bar chart he can request any time dimension (or
interval) of bar. For example, if user would like to work with time
interval in 15 minutes then each bar represents 15 minutes of trading.
Trend tracing is one of keystone
problem. When trend changes preprocessor determines trend and opens new
trade period. Trend can last more than one interval. When trend changes
preprocessor issues trade signal, closes period, and prepares data for
input into receptor layer of associative self-building neural network. Preprocessor should generate trade signals only when
trend really changes. In that case Interpreter will have less problems with
predicting if next trade period is favorable for trade or not.
Main objective of preprocessor
is to work out description of situation at market during open trade period.
One of serious problem was to generate description of the situation that
does not depend on absolute values of prices. It is very important because
associative self-building neural network should store knowledge about
trades based on data from different years. However, prices can differ
greatly year from year. For example, if we compare prices of contracts of
S&P 500 in 1994 against contracts in 2002 we can see that prices in
2002 were in 2 times bigger than in 1994. That is why it was worked out a
set of indicators that track rate of up trade or down trade, levels of support
or resistance, and so on.
Preprocessor has a layer that
consists of a set of sensors. Each sensor corresponds to one indicator,
moving average, or any other information that was selected as a criterion
for trade analysis. Sensors are divided into two types. Qualitative sensors
belong to the first type of sensors, and quantitative sensors - to the
second type of sensors. Each qualitative sensor is confronted with only one
receptor of associative self-organazed neural network (see >>). Qualitative sensor generates only one
signal that activates corresponding receptor. Quantitative sensor is
confronted with a set of receptors. Range of values that sensor receives is
divided into a set of sub ranges. Current version of preprocessor can form
from 3 up to 50 sub ranges. Accordingly, quantitative sensor can combine up
to 50 receptors. In that case receptor becomes activated only if sensor
sends a signal that value corresponds a sub range of the receptor.
an example of qualitative sensor we can use the following indicator: Cn-1
> Cn, where Cn-1 is close value of (n-1) interval, and Cn
- close value of (n) interval. If Cn-1 > Cn
then sensor send signal to receptor that activates it.
an example of quantitative sensor we can use RSI. Suppose its values are in
a range of values between 10 and 100 and set of receptor consists of 4
receptors. In that case sensor sends signal that activates receptor R1 if
value is less than or equal 10, R2 - if value is in a range between 10 and
50th inclusive, R3 - if value is in a range between 50 and 100th inclusive,
and R4 - if value is more than 100.
same time preprocessor processes well known traders indicators and moving
averages and displays them on the screen together with bar chart. It will
greatly helps trader to analyze situation at market and assess advise that
Artificial Foreteller generates.