Machine learning in trading: theory, models, practice and algo-trading - page 1732

New comment
 
Mihail Marchukajtes:

I have been interested in this subject for a long time. I have been messing around with Fourier, wavelets, EMD, SSA. The main problem is latency. All these methods show the past. It is solved by means of forecasting, but it leads to overriding (SSA). If the price is stationary for some time the forecast is good, but how do we know how long the price stays forecasted? The settings in CSSA are quite usual for predicting methods. There is another way to get rid of the delay, it is a phase correction, but because of this indicator has strong deviations after the sharp price movements. I do not think that CSSA hacked all these limitations, they just added a nice visualization of calibration.

 
Just need to check the method. Just myself I will not do it myself, first of all I will do it fiercely long, I remember R here, so the dataframe was doing four hours of unused while I figured it out, there is not really anyone to advise. And what I'll do 100 percent will be a lot of mistakes. And I need to create a script in R, which implements the entire above algorithm and check it in real life. Even if the most difficult parameter TC such as "WARRANTY" will be 3 out of 5 already can earn.
 
Mihail Marchukajtes:
You just need to check the method. Just myself I will not do it in life, firstly I will do it fiercely long, I have remembered R, so the dataframe did four hours of unused until I figured it out, there is no one to advise. And what I'll do 100 percent will be a lot of mistakes. And I need to create a script in R, which implements the entire above algorithm and check it in real life. Even if the most difficult parameter TC such as "GUARANTEED" will be 3 out of 5 already can earn.

this is normal practice, don't be sad, just do it

 
Rorschach:

I have been interested in this topic for a long time. I have fiddled a lot with Fourier, wavelets, EMD, SSA. The main problem is the delay. All these methods show the past. It is solved by means of forecasting, but it leads to overrun (SSA). If the price is stationary for some time the forecast is good, but how do we know how long the price stays forecasted? The settings in CSSA are quite usual for predicting methods. There is another way to get rid of the delay, it is a phase correction, but because of this indicator has strong deviations after the sharp price movements. I don't think CSSA hacked all these limitations, they just added a nice visualization of calibration.

Well of course it will not be able to predict sharp deviations, these will be the very cases of insecurity. But in general, this is one of the correct approaches to the use of frequency analysis in general, rather than this all of your..... minutes they are looking in which to buy in which to sell. HILARIOUS!!!!
 
Maxim Dmitrievsky:

this is normal practice, don't be sad, just do it

Well, yes, so the result would be already by the end of next week, and with me in a year and not a fact, even if I start now. But in general I'm thinking about studying R, the more I work with it, the more I'll get used to its rules. Except for last time I changed the script, I've been running it for about three years, pressed straight+shift+enter and just shut down after some calculations. Hum. :-)
 
So what? The script is spinning. Why get involved if it works?
 
Maxim Dmitrievsky:

Conditionally. 1-e 2 parameters hour and minute, then the quality score and cluster number 1, -1 (you can make more)

1 and 2 parameters can be any (indicator value, etc.), there can be more

You have states now, when the indicators are in some values, you know what strategy to apply. And this will persist for years, with luck

That's the shit I wrote based on your glues.

I just put out a better score, but you can combine

on top of that the caterpillar will work (although you don't need it anymore)


I need to make a primitive mall and put it on OOS

 
So if anyone is ready to advance in this direction I will always be ready to help organizationally. After all, in fact it is necessary to check the hypothesis, if it does not take place, well, okay, but we tried. It would be silly if they did not try, but it is worthwhile. Well again, no one says it will be just perfect work ala grail, but to be profitable I think it will be. For some reason that found the circle does not give me peace :-)
 
mytarmailS:

We need to make a primitive merchant system and put it on OOS

Everything works. We need to come up with floating time windows. Fixed numbers seem limited.

 
All the more that the CSSA algorithm is far from being a secret it was quickly enough cracked and it is not about any prediction, they just added some changes to the algorithm crawler. They build it a little differently. I do not know exactly.
1...172517261727172817291730173117321733173417351736173717381739...3699
New comment