Machine learning in trading: theory, models, practice and algo-trading - page 2293
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
You won't understand until you experiment on sinusoids
Asaulenko experimented on sinusoids, then ran away somewhere... got offended by them
Take a simple 24 period MA and check it. Maybe some other low-pass filterAsaulenko experimented on sinusoids, then ran off somewhere... offended by them
Take a simple 24-period MA scale and check it. Maybe some other low-pass filterhttps://www.mql5.com/ru/forum/86386/page2089#comment_19102675
https://www.mql5.com/ru/forum/86386/page2089#comment_19102675
I don't understand what you're looking for in this way. You need to find a statistically significant deviation from the mean over specific periods, or autocorrelation
it might make sense to go through low-pass filters rather than just MAHs
There is one more obvious possibility for potentially useful application of the forces of cosnics. We are talking about randomness tests in general and NIST tests, for example, in particular. They have everything they like there - cycles, Fourier, patterns, etc. It is true that only binary sequences are studied there, but you can limit yourself to the study of renko graphs for a start.
Alexei, how can you connect the theory of martingale to martingale in forex, to look at it from a scientific point of view on MO + martingale? :)
Alexei, how can you relate the theory of martingale to martingale in forex, to look from a scientific perspective on MO + martingale? :)
You never explained why you need a martin in MO
you still haven't explained why you need a martin in mo
I wrote - a different allocation of space, other opportunities that you do not get by training on one deal
and it's not about increasing the lotPractically, I'm still struggling with a problem where two classes are asymmetrically distributed (one more than 60%) and the grids "burn out" 100% of the time producing one class.
Grids are such a sensitive thing... A small imbalance and they already burn out. Balancing with doubles or removing something from a large class is not good, it seems to me. Especially if the class ratio is 95% to 5%.
This is one of the reasons I switched to wood models (trees, forests, boosts).
Reverse error propagation in networks, also gives its own problems with getting stuck in local points.
The second.
Trees do not have these disadvantages.There are many things you can do, I will not say soda as there are special packages.
Classes should be balanced for the NS. Add examples of missing
It is better to master python.
Balance classes, or redo the metric, which would give more points to a rare class
Thanks. By the way was thinking how to redo the metric, came to mind so - the example if the classes are 0 (30%) and 1 (70%), then for the right value on the reverse pass to give not 0 and 1, but 0 and 0.7 ( or 0.85?) As an option.
But right off the bat I don't know if this is the right way to do it, whether these numbers are right or if I need to play with the activation function, and so on, and right off I haven't found such examples.
Thanks. By the way thought about how to remake the metric, came to mind so - the example if the classes are 0 (30%) and 1 (70%), then for the right value on the reverse pass to give not 0 and 1, but 0 and 0.7 ( or 0.85?) as an option.
But right off the bat I can't figure out if this is the right way to do it, whether these numbers or making magic with activation function and etc, and I could not find such examples.
https://www.mql5.com/ru/forum/86386/page2108#comment_19209601