Quote:
Originally Posted by revvin58
Where do you get 0.8325546 from?
|
radial basis functions is exponential summetric Gaussian functions like exp(-x*x/[2*h*h]).
exp is 2.71... constant like pi=3.14159... . Gaussian functions describe and explain many (all) proceses in our world and uneverse.
for simple we use [2*h*h]=1 or exp(-x*x).
there are many classes of this functions like:
- exp(-x)
- 1/[(1+x)*(1+x)]
- 1/(1+z*z)
- x^alpha*ln(x)
- x*x+alpha*x+betta (polinomial formula) and others.
in graphik its view as old door bell _/\_ .
like binar system 1 and 0 its probability 50% and 50% is equal.
if y=exp(-x*x) and y=50% (blue horizontal level). y=1 y(50%)=0.5
and x=sqrt[-ln(0.5)]=0.83255 (+ and -). for other functions its different.
and parameter of indicator width=5 its depend on this number 0.83255/width like normalization (change this to 1 and see how its look).
its very simple to use indicator: add it to chart window with default settings (euro usd) and see where 10 red predicted balls on price look: for(i=0;i>-10;i--) i0 bigger i-1 bigger i-2 bigger i-3 ... BUY
for SELL i0 less i-1 less i-2 less i-3 ...
and red line in separate window (last 3 or 5 bars) have max or min and invert to other side - this is signal. [if min and rise & red ball is up - BUY] (if max and fall & red ball is down - SELL).
in source you may add new function (for expample copy def to def_new in which close price is replase to open and add in header new second buffer) and then red cross blue from down to up - BUY, to SELL - cross from up to down (see in Rbn++_Double.mq4 test in history use to_past settings).
you can aproximate any discreet function (like MA in each bar just add it instead close price in def).
indicator use neual net in its basic rule like regression with 1000% prediction power.
take profit friends!!!
Radial Basis Network (RBN) - As Fit Filter For Price
NEW UPDate! 04 Nov. 09 
Interesting indicator
