Iteresting topic.
Hm lets see what good advicec i have regarding this:
- Are you sure you are willing to trust a 'black box' system giving you trading advices?
- Define the goals the NN should reach
- Select the needed NN for the problem
- Use stationary data!!!
- Find a way to detect overoptimisation. (Out of sample tests, Monte carlo simulations or similar stuff)
Additionally a few personal experiences:
- Do not try to predict the price, binary decisions worked quite better for me
- Maybe a combination of NN could be a way to go. (1 for trenddetection, 1 for entry signals, 1 for exit signal ecc)
- You probably need to find a way to train thme (and charactersize the results) individually
I was never able to make consistent wins on demo accounts so i never let them go live. Maybe i used the wrong nets or my implementation was not well designed. Or maybe there is no underlying patter which can be exploited in the long run.
You probably need a lot of programming experience and you should be able to write dll's.
I am very interested in the topic, NN where always one of my favorites. Maybe i will give it a shot again. Need to try that java bridge first
This one looks good article but it is in Russian and Mql5 https://www.mql5.com/ru/articles/366 .Maybe moderators might want to translate
into English
This one looks good article but it is in Russian and Mql5 https://www.mql5.com/ru/articles/366 .Maybe moderators might want to translate
into English
Or you could try
which does translate webpages to English, sort of.

- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
Hi,
I have read most articles on NNs, and I am fascinated by them. :)
I am looking more at analysing the major fx pairs going back 10 years and trying to find any patterns.
Has anybody used NNs for candlestick analysis, if so; how did you get on and can offer any advice?
davefx