Neural Networks for trading - page 4
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
Then respect for that dude, because it's an arduous project you have there. I'll be glad to be informed of the results you got by doing it that way because it's an extremely purist approach.
The idea is that it would analyze the last 100 candles OHLC and find patterns by experimenting and learning thru many many (many many many ...) deals which ones suits the market. There's probably a system of reward somehere also.
I understood well, did I ?
Like putting all together the last 100 open, high, low and close... Those are 400 inputs at least (if you don't add any other thing).
No idea dude, it's Marco's concept. Wait his answer to get enlightened
No it's not about 100 or 400 candles.
It's about as many as you can get.
If it was 400 then it would not take long to train the network.
The data for these tests started in 1971 up to 2015 (EURUSD) and 1996 - 2015 (AUDUSD) in the graph i posted.
I do not know how many candles that is but i know it's not 400 maybe more like 400 Million ? or 400.000 input datapoints.
I would have to check the csv files.
If you look at that graph above the black dots are the historic datapoints.
Yes the activation can be Sigmoid or ReLu rectifier (with dropout) depending on the application if its a classic perceptron or some more advanced model.
But that is trial and error for the investigator because he is aught to try many to come up with the best solution, eventually.