Machine learning in trading: theory, models, practice and algo-trading - page 3249
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
But I do in python and calculate the correlation for all possible pairs at once, then choose from that.
The most important thing here is speedThere must be an analogue of this in Python. Then it should be fast.
Saw something similar here.
It looks similar on the screen, but the length of the pattern is very high, because M1. It will probably show something interesting on the hour-markers, since it was found.
And this problem
seems to be solved.
But it's not mining (overkill) there, of course. Although not far away.))) I saw it too
I was originally interested in how to search for patterns in multidimensional arrays without MO. So far I haven't come up with anything better than cramming all the dimensions into one and calculating via correlation (kinda fast). I guess sometimes the values need to be normalised so that they are not too different.
There must be an analogue of this in Python. Then it should be fast.
There is, but when the number of signs (indicators) grows, it is still not very fast.
The 3980 implemented Conjugate methods for the types complex, vector<complex> and matrix<complex>. They perform conjugation for complex numbers.
They also added processing of ONNX model output of Sequence of maps type. The functionality of ONNX Runtime has been seriously improved.
There is, but with the number of signs (indicators) growing, it is still not very fast.
I don't remember, but the complexity of the algorithm is definitely less than O(N^2). I think it is not higher than O(N*log). That's why it's not quite clear the noticeable slowdown when the signs grow.
There is a two-way stick: more features, fewer samples - lower statistical significance.
I guess sometimes the values need to be normalised so that they are not too different.
It can be a mess without normalisation.
Forum on trading, automated trading systems and testing trading strategies
Machine learning in trading: theory, models, practice and algo-trading
fxsaber, 2023.09.21 16:19
Then it is necessary to bring indicators to some unified parrots. Even if the indicator is the increment at different intervals, otherwise a strange correlation will come out.
Forum on trading, automated trading systems and testing trading strategies
Machine Learning in Trading: Theory, Models, Practice and Algorithm Trading
Maxim Dmitrievsky, 2023.09.21 15:59
that example from the tester had a length of 9 (increments of different periods)
it is not at all clear what the correlation shows without normalisation. What if. this skipped, seemingly logical step leads to a good result.....
I don't remember, but the complexity of the algorithm is definitely less than O(N^2). I think it is not higher than O(N*log). That's why it is not quite clear the noticeable slowdown when the signs grow.
There is a two-way stick: more features, fewer samples - lower statistical significance.
Well, I'm also counting all possible pairs at once. There's still a lot of input data I want to try. That's fine. It's just that in STUMPY there is an opportunity to approximately count and then refine. You get a noticeable acceleration, plus paralleling and on the GPU. I will probably switch to that package completely.
That's where mush can come out without normalisation.
If it's non-standard returns, like this, .it's not clear what the correlation shows without normalisation. What if... this skipped, seemingly logical step leads to a good result....
I'll analyse it later, I'm not ready to comment on it yet, I just wrote this calculation yesterday.
I'll analyse it later, not ready to comment yet, just wrote this countdown yesterday
I think that the correlation will be influenced by the largest numbers in terms of abs value. For example, a volume change of 10000 and 10100, and on their background a price change of 0.00040 and 0.00400 will be microscopically small and will have little effect on the correlation of the whole set. I would do a normalisation to test this hypothesis.
I had to think hard to figure out how to explain it on my fingers.
The answer is naturally as simple as an egg.
There was a very cool thread in which I have a post:
Failure at the end of trading is because the test ends on unclosed positions - MQL4 and MetaTrader 4 - MQL5
10 years ago I already made such programmes, and you have no such profits here, not even in the tests
And in real life this EA has been trampling around zero and no growth, absolutely.
So the conclusion is obvious
real trading will never be like a tester's grail.
which, among other things, is driven by the MOSHKA
But in order to make money consistently, and more and more, it is necessary not to forecast, not at all.
you need to understand and figure out the algorithm, i.e. how any market with a floating price works.
the algorithm is the same, 100%
and only after that you'll have an indication like this, which will emphasise the possibility of making money.
that's the whole point of it, after that you don't need it at all.
;)
lying?
No,
I'm attaching the figures from real life too.
and the balance with a minus, definitely not a demo ;))))))