Build Neural Network Indicator in MT4 using Neuroshell

I am thinking how to implement this;
- Example I have data RSI, CCI,stoch%D and StochSlow%D as x0,x1,x2,x3
- I have ideal trading signal, let say 1 for buy, -1 for sell and 0 for hold as y
- Let say I have 10000 rows of data
- when i try to put into eureqa .. y=step(f(x0,x1,x2,x3)) the function resulted is f(x0,x1,x2,x3)=0
probably because most of data signal is Hold (0) values.

Should we remove 0 data from the training and kept only signal consist of buy (1) and sell only (-1)?

Any idea?
 
So far I am happy with my trading result..but all the details in myfxbook are kept private.

Happy trading...
 

Attachments

  • trading equity 21 Apr 2011.png
    trading equity 21 Apr 2011.png
    37.7 KB · Views: 630
  • trading equity 22 Apr 2011.png
    trading equity 22 Apr 2011.png
    37.8 KB · Views: 507
Someone posting in other forum..
=====
Am I ??

Desperate trader.....
Trading with 99% lost
Profitable trade only by luck..
Hunting indicators but not using consistently..when fail then dumb it..
Using trade signal based on friend information ..even though the friend mostly lost and rarely get profit
Subscribe trade signal ..but always be cheated..
Believe that using EA will make profit..but actually no
Buying EA or software ..but never get profit and only get cheated and loss money..
When EA disappointing .. dumb it or share it to the forum due to angry to the creator..and someone else can get loss too
Want to get profit efficiently ...then download free or cracked version only solution..
Asking someone to give the licensed copy....but nobody give and not sure that he has or not...
Always hunting profitable EA/software or cracked version ...with no guarantee
Money management just only motto..never implement it during trade
I always using emotion during the trade ...kept lost open wider and close the profitable trade soon (even a little)..
Holly grail trading strategy still on dream ..but never come true...

The only best friend still remain the same..... no money - no profit - no license ....and losses still resist
Wake up man...nobody will give you money free ...without hard work even with the best trading software in the world you will still get lost .. and because you never change your habit...

====
 
Stay away to trade during long holiday..the spread become wider

Gold 4.25 (425 pips) and silver 1.26 (126 pips)...normal one is only 50 to 65 pips ...
 

Attachments

  • Easter holiday spread.png
    Easter holiday spread.png
    19.1 KB · Views: 576
  • Easter holiday spread2.png
    Easter holiday spread2.png
    9.8 KB · Views: 1,612
So far I am happy with my trading result..but all the details in myfxbook are kept private.

Happy trading...

arry your returns are good, but is your system biased only for longs (the fact that gold and silver are in bull phase now since recession) what about short trades winning .
 
Hi Supremegizmo,

See my sell trading sample..it should self explanatory...

Can you show your own trade also ..it would be nice if you can show it..
 

Attachments

  • Sell trade.png
    Sell trade.png
    52.9 KB · Views: 847
Here another way to analyze market..using Elliott Wave

Nice to try Elwave..
 

Attachments

  • silver using elliot wave.png
    silver using elliot wave.png
    186.9 KB · Views: 1,741
  • gold using elliot wave.png
    gold using elliot wave.png
    168.6 KB · Views: 2,142
Hi Supremegizmo,

See my sell trading sample..it should self explanatory...

Can you show your own trade also ..it would be nice if you can show it..


Ok Arry I will post my trade tonite right now my PC is doing some hard core testing. (y)
 
Hi Supremegizmo,

See my sell trading sample..it should self explanatory...

Can you show your own trade also ..it would be nice if you can show it..

Arry here is small example of one of my strategy.. fingers crossed it keeps workking. This is example when the pair is consolidation it makes go money. I dint mention I lost 200 pips on last week when gbpusd broke to upside :cheesy:
 

Attachments

  • gbpusd.bmp
    1.4 MB · Views: 469
  • gbpusdresults.bmp
    1.2 MB · Views: 444
Supremegizmo,

I think you may add stop loss in exit strategy..possible to add higher time frame signal for exit confirmation..
 
Supremegizmo,

I think you may add stop loss in exit strategy..possible to add higher time frame signal for exit confirmation..

Arry

I left the neural net to select the exit threshold. I mean the long/short entry/exit

thresholds are selected by neural nets and it not interfered by me. :cheesy: Its ok atleast its giving me some decent returns

No that I know nsdt is truly amazing if one uses practical methods with a bit of '3D' imgaintiaon then he can create strategies(y)
 
Take benefit for the USD low interest rate ...

Till when..I do not know...

Happy trading
 

Attachments

  • 2nd month equity growth update.png
    2nd month equity growth update.png
    27 KB · Views: 1,075
Hi All,
I do little observation and found that indicator created from NS2 recurrent network (Jordan-Elman) tends to repaint. It might be as a result of recurrent network behavior that memorized input sequence. Can anybody confirm it? If it really true, is there any idea so that we can still take an advantage of this kind of recurrent network model?
 
Hi Sgr,

Welcome to the thread..

There are few possibility that making your indicator repaint, again it may depend on what inputs and output you are using in your NN.

Another possibility is during presenting the trained net as indicator, I prefer to use the previous completed values to draw indicator instead of current Close price which is not completed yet. My self, using previous completed indicator value to generate the trade signal, if you are using current indicator as trade signal, most probable the signal will repaint.

Anyway I will try to train a net using recurrent network and compare to other nets.
 
Thanks Arryex,
Currently, I use no-repaint inputs, and use previous completed values to draw indicator. I compare two networks, one with recurrent network, and the other using Ward Nets, by using the same indicator code and inputs.The first one is repaint, and the second one is not repaint. During my observation, it is easier to get higher R square and correlation coefficient using recurrent network rather than ward nets.
 
Can you post your data, hence we can use the same problem reference? I could not work based on guessing..

Thanks
 
Top