Neural networks programming with prorealtime
Forums › ProRealTime English forum › ProBuilder support › Neural networks programming with prorealtime
- This topic has 126 replies, 8 voices, and was last updated 1 year ago by MobiusGrey.
Tagged: data mining, machine learning
-
-
12/21/2018 at 2:08 PM #8756312/21/2018 at 3:14 PM #8756412/21/2018 at 4:22 PM #8756701/24/2019 at 1:20 PM #89567
Hello Nicolas,
I like your idea! Could you complete this code with also short trades?
Thank you 🙂
04/27/2019 at 2:21 PM #97256Hello! Is there any update about this system? I don’t understand it too much, and I think this is the future but seems interest is decreasing in the forum.
Thanks a lot.
04/27/2019 at 4:01 PM #97258I don’t understand it too much
I never understood it at all 🙂 but I managed to turn the code into Auto-Systems, many of which I have been running since … attached are results of a few.
1 user thanked author for this post.
04/28/2019 at 5:14 AM #97278WoW! That’s awesome @GraHal! What indicators are you using in profitable strategies? Results are very good! Is real trading or paper?
Thank you very much for your feedback!
04/28/2019 at 9:40 AM #97282Results are very good!
When you consider that the DJI has risen from 21,000 ish to 26,500 since Dec 18 (when the Systems were set running) then it’s not so good?
What indicators are you using in profitable strategies?
For the most profitable System (I show above) then all I did was add below.
123456789101112131415161718192021If Output1 > 0.7 and output2 < 0.7 ThenBuy at MarketEndif//and ADX[14] > 19//If barindex - tradeindex > A351 and POSITIONPERF < 0 Then//Sell at market//endifIf Output2 > 0.9 and output1 < 0.9 ThenSellshort at Marketendif//and ADX[14] > 12If LongonMarket thenSET TARGET PPROFIT 520SET STOP PLOSS 60 //60EndifIf ShortonMarket thenSET TARGET PPROFIT 20 //50SET STOP PLOSS 20 //25EndifAll credit and thanks must go to Leo and I’m sure his code is capable of giving better results … anybody brave enough to show what you got?? 🙂
04/28/2019 at 10:01 AM #97284Analysing results further, what is significant is that the winning trades could hardly be described as a a rocky road / stressful journey … look at the low MAE values against gains of 520.
The potential of Leo’s code needs to be unlocked, but I don’t understand it enough to make the right key! 🙂
EDIT / PS
The screen shot below is from a Backtest as MAE isn’t shown on Forward Test Detail Report (is MAE shown on Fwd Test version 11 … I hope so!?).
04/30/2019 at 8:11 AM #97386Thank you very much @GraHal for your efforts. I hope these days I’ll have time enough to study this system and could talk about it later.
Kind regards.
08/30/2019 at 5:57 PM #105936Hi Leo it’s a great job, even if I arrived late but this topic attracts me a lot.
To give support to your work you might be interested in this link https://www.mql5.com/en/code/9002,
you can also find files for MT4 that could be converted with the help of experts.
Please if you have made further progress I ask you if you can post it without commitment, thanks and greetings
08/30/2019 at 9:37 PM #105960Hi all, thank you very much for your comments,
I tried my best to improve performance but it is very hard coding a neural network… and I do not improve any further. I also explore other technics of machine learning like clustering. In one of my codes the memory issue with prorealtime only allow me to do a loop of learning with 3 epochs (a for-next loop only 3 times!). I got only a multi-layer perceptor of 5 inputs with like 6 neurons… when we talk than a “simple” professional neural network has dozens of inputs with hundreds of neurons and several layers… so I got very disapointed with prorealtime.
Now I concentrating in learning Python from zero, I’m afraid to say I do not code anymore in PRT. For artificial intelligence there are many libraries, so most of the work is done already. Thanks for sharing your page Coscar, thanks for your comments GraHal ( it encourages me a lot to keep going) and thanks a lot for all the community as well.
3 users thanked author for this post.
09/04/2019 at 8:29 PM #106529Hi all,
here is the best approach I got
Neural Network v2.2.1
Here I attached the file,
remember to upload at least 15000 bars.
I wish all of you a lot of success.
7 users thanked author for this post.
01/10/2021 at 11:08 AM #157175I wish all of you a lot of success.
I wish i had read this thread a long time ago . Great body of work here and i wish i could understand more of it . I will put some time into learning about neural networks as it has an appeal .
We all have a Neural Network !
I am certain your hard work has paid of by now . Kudos to you Leo
AND yeah yeah blah blah dont do this do do that . signed the narcissist in the boat
01/10/2021 at 11:33 AM #157178AND yeah yeah blah blah dont do this do do that . signed the narcissist in the boat.
Follow the few simple rules and the narcissist in the boat wont have to keep editing your posts and reminding you of them. Other forum members seem to manage to follow them.
Rules such as:
- Always be polite and courteous to others.
-
AuthorPosts
Find exclusive trading pro-tools on