ULE price events
Hello, Group
This week you have mentioned Price events in ULE module. For statistics
of technical analysis indicators it is better to use Chaos theory
module, I will explain how it works soon.
Price events in ULE - this is another and long story. Let me explain
what is that. Initially these events have been developed for a special
category of Neural Networks - the model that forecasts NEXT DAY price
movement. In other words analyzing price history the model forecasts the
price movement for tomorrow.
The most appropriate math tool for this task is auto regression (AR) model.
http://www.timingsolution.com/TS/Articles/anti_trend/index.htmand these Price events in ULE have been developed for this task. De
facto these events represent FUZZY logic structure of the price chart
information.
You can look at this article about fuzzy logic math that has been
written 12 years ago already:
http://www.timingsolution.com/TS/Articles/Fuzzy/After that in 2007(6)-2009 years we conducted a massive backtesting for
auto regression models, we tried to get next day forecast. The math
behind this model is very powerful and very universal - NN based on
fuzzy logic.
But we did not get any results, we could not get next day forecast. I do
not tell that this is impossible - it was recognized that this task is
much bigger than we expected, this is a huge task, this is "digital
fortress". Any parameter here can totally change the picture, while
taking that parameter into account needs calculations, calculations,
calculations .....
It was recognized that to go further we need "technological
breakthrough", the computers that work much more faster, thousands and
thousands time faster, may be parallel computing or even quantum
computers (it looks like they should appear soon).
So I keep these prices for this task. I don't know when we will continue
working in this direction.
If sometime we can get next day forecast, it will be possible to make
the next step - building fuzzy inference machine. This is a system that
allows to gather KNOWLEDGE from fuzzy NN. This is a very interesting and
appealing feature, something like that:
1) using very powerful computing, we finally get a next day forecast
model;
2) special procedure extracts/gathers knowledge from this NN i.e. it
will be pieces of knowledge like this "if RSI <32 AND First quarter of
Moon AND RSI 5 days ago < 25 => next day price goes up". In other words
it will be possible to obtain the set of rules understandable by traders.
NN works as "black box", we don't know what is going on there. This
technology allows to convert the knowledge hidden in Neural Network to
some set of rules understandable by anybody.
This is really cool, but before that we have to get next day forecast
model. Without positive results there we cannot go further.
Best regards.
Sergey.
