A quick and free library for MT4, much to the delight of neuralnetworkers - page 19
![MQL5 - Language of trade strategies built-in the MetaTrader 5 client terminal](https://c.mql5.com/i/registerlandings/logo-2.png)
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
I was testing EUR/USD H4 from 2006 to 2008 (shifting a month forward after each optimization) and then substituting values for 2009, month by month, 10 times for each month, as they change every time . As a result, in 7 out of 10 cases I got a positive result, and so I did it for 2009. Where is the hidden clue? Why cannot it be used? I do not understand half of your speeches, let me explain in simple terms.
1) Which Expert Advisor are you talking about?
2) Describe the method of optimization clearly. Honestly - I have read it five times and do not understand it once. ))
I'll start at the beginning...
Tested on EUR/USD H4 with genetics.
Period of optimization is 2 years. I used this optimization for 31.12.2006-31.12.2008 in January 2009 to pick SL (like a theoretical trade). Then for February 2009, I used the data from 31.01.2007-31.01.2009. And so the whole year 2009 and the beginning of 2010. The SL obtained for the January 2009 trade was 84. I used this value for the period 10 times, to make sure of a good probability of profit. I made profit 7 times out of 10. This was the case during the whole year. And if the loss was quite small, the profit skyrocketed.
Thank you for your feedback. I've never been much into EAs, so I'm surprised.
Talking about FANN-EA https://www.mql5.com/ru/code/9386#25259
I'll start at the beginning...
....................
Thanks for responding. I've never been much into EAs, so I'm surprised.
I also tried it again from the beginning and did not find anything surprising. On the test, in January 2009, the usual plum....
But do not forget that a neural network Expert Advisor is a separate class. And even if you do everything "the same" the results will be different, because neurons.... weights.... etc.
If it's interesting to get to the bottom of the reasons for the differences, let me know, I'll write an algorithm of our actions, and then we can talk substantively.
Please write what improvements you have come up with to use this EA and library
1.) The trailer contains standard FANN-EA, the only difference: the external variable extern bool TrainOnTesting = false; has a switch to train the network only during optimization or during testing too. Rename the Expert Advisor to be different from yours and use this variant.
2) Clear the ANN folder, save a copy if needed.
3) Optimise over two years (2007-2008) as you did before. (I have a 30 minute run with genetics). then choose a set you are happy with and run on January 2009. The results will not change from test to test because TrainOnTesting = false. It is necessary to fix the weights of the trained network.
Report the results...
started from September 196.
October 132
November to March 164.
and then I ran September again and it popped up 164.
What should I do with the neuron before testing it in real time on the demo?