Machine learning in trading: theory, models, practice and algo-trading - page 2234
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 have a classifier on two-dimensional ultra-precise keras networks. The input is a graphical representation of the quote, which is sawed by matplotlib. In my tester, which takes into account the spread on each position, everything is fine. Test for 2020.
I tried to pair it with mql using keras2cpp but it did not work. So I decided to use file. I have to use the file, it is slow enough)) and it will not change anything.
i got a strange looking mirror image of the balance graph in both directions from the center )
We need to do something about the spread. If you turn it off in the tester, it learns well and for 15 min. Not the best, just as an example. I do not believe that at 15 min it is so important that it kills all profits. It seems to me that somewhere there is an error, it is not correctly laid in the mb model.
I think there's a fairy tale here at 15 min (no spread).
)
I also tested on 15 minutes and had the idea to add one more condition to your random trade generator, do not trade. Include it on some condition, such as profit size is about the spread, or the deal is long in the drawdown, or both, in short, to add to the model some real trading circumstances through this condition. I want to add more, but there are only 24 hours in a day)))
suspiciously mirrored balance graph, if you look from its center to both sides )
Hah, exactly) This graph is obtained by the tester with animation in real time, which I posted here earlier)) I don't believe it yet and I think the MT tester will put everything in its place.
Seriously... this weirdness is breaking my brain. Without the spread, it works well on the track/test.
With the spread, it works well on the track, but it breaks on the test. What's so different on the test that the spread tick by tick prevents it from turning into a profit...
Looks like some kind of bug in the logic
Seriously... this weirdness is breaking my brain. Without the spread, it works well on the track/test.
With the spread, it works well on the track, but it breaks on the test. What's so different on the test that the spread tick by tick prevents it from turning into a profit...
sounds like some kind of bug in the logic.
Count the number of trades and multiply by the spread. And it's possible that the spread is not taken into account correctly in training.
The world is random. He is being treated (his health is good)) ) I have a professor of mathematics from the president's financial school. Today I found out that he has been practicing on Python in Keras for five years on dark matter, ECG and various other things)))
It seems that nuclear physicists have CERN ROOT - a very cool system, but hard for beginners. It uses C to work on the command line. However, with respect to the matstat it seems a bit poor in comparison to R. By the way, they use MO there quite often- apparently Schrödinger and the cat can no longer cope)
The world is random. He is being treated (his health is good)) ) I have a professor of mathematics from the president's financial school. Today I found out he had been practicing on Python in Keras for five years on dark matter, EKG and other stuff.)
Financiers EKG and dark matter research...
Another case in point. An acquaintance - an economist at Moscow State University graduated, for her diploma studied the dependence of the social survey on search queries in Yandex. (This is the business of sociologists.)
Isn't that strange? Financiers and economists do non-core stuff.
With finance it seems to be a waste of time. Like us here, they've been struggling for years with zero results.
The nuclear physicists seem to have CERN ROOT - a very cool system, but hard for beginners. It uses C for the command line. However, with respect to the matstat it seems poor in comparison with R. By the way, they use MO there quite often- apparently Schrödinger and the cat can no longer cope)
Schroedinger can not cope for a long time, today I read his review on the dissertation about numerical simulation of nonlinear spin waves in graphene structures on matlab, of course they use Schroedinger, but already noting that he can not cope))))
It seems to be a hobby at home) In general for a long time surprised, the academy is financial, and talk about quantum computers, CERN, Dubna) But today surprised just) The child (my) in graduate school, he asked a question, and the network in Python writes?) I asked him a question).
Financiers EKG and dark matter research...
Another case. An acquaintance, an economist at MGU, graduated with a degree in economics, and researched the dependence of social surveys on search queries in Yandex. (This is the business of sociologists.)
Social statistics is more ancient, and matstat is the basis of the basics. And among sociologists matstat few people know at the necessary level apparently.