You are right we can't use a feature value from moving averages for
forecasting the signal.
We must find a solution for this.
Than you for your obsevation.
Quote:
Originally Posted by adimsh
http://www.forex-tsd.com/211176-post207.html
The image shows 100% acuracy in classification, this cause doubt, there is no solution that will give 100% acuracy, there must be some thing wrong.
To barnix, I have a question:
In your code for "5_bar_SVM_input.mq4":
Code:
string Line1(
int nb1,
int FW,
int Price1,
bool train,
int pr1,
int i
)
{
int sig0;
string ss0;
string ssig0;
if(train)
{
sig0 = iCustom(Symbol(),0,"5_bar_sig",
nb1,
FW,
0,
i+0);
}
else
{
sig0 = 0;
}
string x1=DoubleToStr1(iMACD(Symbol(),0,FastPeriod,SlowPeriod,SignalPeriod,Price1,MODE_MAIN,i+0)/Point,pr1);
string x2=DoubleToStr1(iMACD(Symbol(),0,FastPeriod,SlowPeriod,SignalPeriod,Price1,MODE_MAIN,i+1)/Point,pr1);
string x3=DoubleToStr1(iMACD(Symbol(),0,FastPeriod,SlowPeriod,SignalPeriod,Price1,MODE_SIGNAL,i+0)/Point,pr1);
string x4=DoubleToStr1(iMACD(Symbol(),0,FastPeriod,SlowPeriod,SignalPeriod,Price1,MODE_SIGNAL,i+1)/Point,pr1);
you are having a signal on bar "i" using iCustom, then you are registering a feature (which is MACD) calculated at bar "i" too.
In my openion, you could not use a feature value at the bar you want to predict the signal at, you should calculate iCustom for "i-1".
Thanks for your great work and knowlage.
Regards,
|
|
07-17-2008, 01:35 PM
Linear Mode
