Machine learning in trading: theory, models, practice and algo-trading - page 1286
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
Does Perervenko have any results?! He does not and cannot have them by definition.
I am not going to work with ZZ - it absolutely contradicts the very meaning of time series, because the concept of "time" is lost.
Perhaps I am too fixated on the theory of stochastic processes, where time is the fundamental variable. Let someone change my mind - show me the account monitoring for the strategy with ZZ.
Please explain what do you mean by "time", some kind of couplings with harmonics, like furies, or harmonics as features?
For the billionth time, I attach a book that is invaluable when using neural networks and scaffolding.
From it even a Papuan understands that you have to have a stationary BP at hand, where price values are obtained at equal integer time intervals. That's it.
How to do this? Do I know?! Aliosha, for example, collected values every 1 second, and Doc - I don't know, but I do know that Doc managed to get BP to a stationary form and he stupidly predicted the sign of the future price increment.
For the billionth time, I attach a book that is invaluable when using neural networks and scaffolding.
From it even a Papuan understands that you have to have a stationary BP at hand, where the price values are obtained at equal integer time intervals. That's it.
How to do this? Do I know?! Aliosha, for example, collected values every 1 second. I don't know, but I do know that Doc managed to get BP to a stationary form and he stupidly predicted the sign of the future price increment.
But nobody works with price as it is, we predict future returns by past, sign and magnitude(volatility), log returns are stationary enough if you level up the re-distributive heterecedenticity. I'm often astonished by crying about non-stationarity, as if somebody forecasts the price directly, it's usually theoreticians who never tried to forecast or build TS, they just hear some nonsensical phrase about "non-stationarity" and repeat it everywhere...
If you're trading intraday with a couple dozen trades a day, even a hundred, a minute will do.
But no one works with the price, we predict future returnees by past, sign and value (volatility), log returnees are stationary enough if you align the reperiodic heterecedentricity. I'm often surprised by crying about non-stationarity, as if somebody forecasts the price directly, usually theoreticians who haven't tried to make a forecast or build TS themselves, they just hear some scientific phrase about "non-stationarity" and repeat it everywhere...
Well, I can see that you're already a know-it-all. Then why do you ask? Pererwenco is a bit of a mouthful...
Well, I can see that you're good at it. Then why do you ask? You're bringing up Perervenko...
He can't be deceiving everyone together with the FA, it's a conspiracy... I am scared...
He can't be deceiving everyone along with the FA, it's some kind of conspiracy... I'm scared...
:)))
He can't be deceiving everyone along with the FA, it's some kind of conspiracy... I'm scared...
This is the Rpidemic. You can go there and not come back, doing reviews on hundreds of packages. It will pass like everything (new) in life.
Here Alexander always writes the same thing, in different words and so to speak and their derivatives, to get to the children of all kinds. C is stability.
But no one works with the price as it is, we predict future returns based on past ones, sign and magnitude (volatility), log returns are stationary enough if you align the reperiodic heterecedentricity. I'm often astonished by crying about non-stationarity, as if somebody forecasts the price directly, usually theoreticians who never tried to forecast or build TS, just heard a scientific phrase about non-stationarity and repeated it everywhere...
I totally share your indignation!
But you should also understand us - we "shouted" only because we were waiting for you to come and show us all how to predict the price through returns and log returns.
Burn!
I also made the return from training as a probability, not as a class number. (which limits the application of the forest for classification to only 2 classes 0 and 1).
Nothing to say about trading/testing results, I only see probabilities so far. And still testing permutation. I have not progressed any further. One recalculation via deletion takes 6 hours. ((( Now evaluation on valid want to compare. I think the train is still overfitting - and I was evaluating the importance for fitting.
If the limiters on the depth and the number of examples in the sheet are in modern scaffolding, apparently it makes sense. That's why I'm doing it myself.
This is what it means to be a real programmer, not like us theorists.
For the billionth time, I attach a book that is invaluable when using neural networks and scaffolding.
From it even a Papuan understands that you have to have a stationary BP at hand, where the price values are obtained at equal integer time intervals. That's it.
How to do this? Do I know?! Alyosha, for example, collected values every 1 second, Doc - I don't know, but I do know that Doc managed to get BP to a stationary form and he stupidly predicted the sign of future price increments.
A_K, can you stop talking nonsense and refer every time to Kolmogorov's article, which you yourself don't understand).
By the way, what is it about, the article? After all, the main thing in the article is not stationarity and counting at equal "integer time intervals". In the article, it's just the initial assumptions made to prove that... By the way, to prove what exactly? Wrong, not predictive capability at all. Then what?
The article says nothing about the fact that prediction is impossible for non-stationary processes, for processes not in "integer time intervals" but with some other. All this, by the way, is also possible, with just as much success.)
In general, your endless claims about the possibility of predicting stationary time series + through "equal, integer ...", other than nonsense can't be called anything else. And Kolmogorov's article has nothing to do with this nonsense - it's about something else.