Welcome to Forex-TSD!, one of the largest Forex forums worldwide, where you will be able to find the most complete and reliable Forex information imaginable.
From the list below, select the forum that you want to visit and register to post, as many times you want. It’s absolutely free. Click here for registering on Forex-TSD.
Exclusive Forum
The Exclusive Forum is the only paid section. Once you subscribe, you will get free access to real cutting-edge Trading Systems (automated and not), Indicators, Signals, Articles, etc., that will help and guide you, in ways that you could only imagine, with your Forex trading.
Elite Section
Get access to private discussions, specialized support, indicators and trading systems reported every week.
Advanced Elite Section
For professional traders, trading system developers and any other member who may need to use and/or convert, the most cutting-edge exclusive indicators and trading systems for MT4 and MT5.
Well, I hate giving you ready solutions considering I failed to come up with something interesting and useful. I hope you could try something different. I mentioned before the learning curve is very freaking steep, you have to read a lot, all info is available on the net including this site. There are threads devoted to this issue.
Briefly, since you built your normalized inputs now you have to build your NN and teach your NN by those inputs. NN consists of group Neurons which produce some kind of Output. Keep in mind Output also has to be normalized and preferably should lay between 0 and 1. So, having some function PNN which called on each new bar -> this function applies your Inputs on N bars back, builds neurons and gives you normalized Output. So, again your Output would be different on each bar in correlation of your Inputs processed N bars back in history. Now, you could have two external parameters in your expert Threshold_Buy and Threshold_Sell and if Output>Threshold_Sell you sell, if Output<Threshold_Buy you buy. In MT4 optimizer you could optimize those 2 parameters to have the best values (on determined period of time), then you could test your expert on period which is out of optimization period. If everything still look okay at that point then you are on the right track.
Keep in mind that's what I was doing.....
Orest
Quote:
Originally Posted by jasondavis87
This is what i currently have:
below -80 is a sell signal and above 80 is a buy signal
the Aqua line is a Net w/ the following inputs:
3MA's(8,13,21) PSAR & Bollinger Bands
The Lime line is a Net w/ :
CCI Stochastic, RSI, %R, DeMarker
The Red one is :
MACD, Momentum, Gator
All values are normalized using the same method
The lines are the outputs from the network after training
kool ... i understand ... i've been trying to get a working solution for the last 3 months now with no concrete success.
I had another idea ... i was thinking the earlier today that my problem might be this: when training i skip bars ... IE i train the indicator based on the highs and lows of a trend ... this might be messing up the results when i test it because i test on every bar ... i was thinking ... its not trained to do every bar ... which would put my zigzag theory out the door ...b/c it lags ... just a thought ... does this sound like it could be a problem?
Here's another thought ... if we're training the indicator on a zig-zag which only goes up and down ... how do u get a "Do nothing" in there ... zig-zag is either up or down
I think I understand what you are saying about the missing info in the bars - - if there is a range bar type available, then the price and volatilty might be more accurately reflected for NN analysis.
Range bars would fix the range per bar and more accurately reflect activity of each currency pair since fast and radical price range moves in a shorter amount of time - - like an implied time and sales of sorts...
Linear Mode
