Machine learning in trading: theory, models, practice and algo-trading - page 548

 
Mihail Marchukajtes:

I will support the topic. I use Amazon services, but their model builder is not so good. In any case, I could not build more or less5 quality model. Although maybe I did something wrong, but there are not too many settings there. I will now google it...


start with this article :) you can learn a bit of python as well... and the link above to the website of the dude there all is chewed up. python is the easiest language to learn.

http://www.blackarbs.com/blog/time-series-analysis-in-python-linear-models-to-garch/11/1/2016

i will copy and paste it for google soon, it is really handy

Time Series Analysis (TSA) in Python - Linear Models to GARCH
Time Series Analysis (TSA) in Python - Linear Models to GARCH
  • 2016.11.08
  • Brian Christopher
  • www.blackarbs.com
So what?  Why do we care about stationarity?  A stationary time series (TS) is simple to predict as we can assume that future statistical properties are the same or proportional to current statistical properties.Most of the models we use in TSA assume covariance-stationarity (#3 above). This means the descriptive statistics these models predict...
 

GARCH gives an error, everything else works

Notepad

 

Google service took a glimpse. This is, as I understand it, a Jupiter laptop. You can run it locally. Yes, it is a handy thing. But I still prefer the IDE. I use a light IDE Visual Studio Code.

 

https://it.mail.ru/video/playlists/ Courses from Mail Ru, including courses on machine learning and data analysis.

 
Grigoriy Chaunin:

Google service took a glimpse. This is, as I understand it, a Jupiter laptop. You can run it locally. Yes, it is a handy thing. But I still prefer the IDE. I use a light IDE called Visual Studio Code.


It's a variant of Ipython, it's convenient for research... and it's really convenient, and then it's easy to be converted into a regular .py

 
Maxim Dmitrievsky:

GARCH gives an error, everything else works

notebook


I do not understand the model itself: it should consist of three parts: arima (for trend), ARCH (for volatility and there are many of them), and distribution. In the text, the coefficients for ARIMA, but in the formula, what do they refer to? For the arch, too, we need to specify similar numbers. All in all, everything is not clear - I don't see any way to steer through the details.

According to the presented material, it looks like a toy.

 
SanSanych Fomenko:

I do not understand the arch model itself: it should consist of three parts: arima (for trend), ARCH (for volatility and there are many of them), and distribution. In the text, the coefficients are for ARIMA, but in the formula, what do they refer to? For the arch, too, we need to specify similar numbers. All in all, everything is not clear - I don't see any way to steer through the details.

From the material presented, it looks like a toy.


I'm still focused on learning python itself, so I haven't looked into it in detail... here's the documentation on it https://pypi.python.org/pypi/arch/4.0

a lot of packages there are direct analogues in R, so there shouldn't be much difference

the fit() function specifies whether the row is stationary or not

this article is more of an introductory one... for example my arima didn't work until I put bfgs instead of lbfgs solver... and garbage won't fit at all... maybe the python version is different, go figure it out :) you'll have to study every lib

arch 4.0 : Python Package Index
  • pypi.python.org
ARCH for Python
 
here with
SanSanych Fomenko:

I do not understand the arch model itself: it should consist of three parts: arima (for trend), ARCH (for volatility and there are many of them), and distribution. In the text, the coefficients for ARIMA, but in the formula, what do they refer to? For the arch, too, we need to specify similar numbers. All in all, everything is not clear - I don't see any way to steer through the details.

From the material presented, it looks like a toy.


Here is an article and notebook from quantopian, maybe it is clearer there

I will hang out at that resource for a while, see what people are doing, maybe there is something interesting

https://www.quantopian.com/posts/quantopian-lecture-series-arch-garch-and-gmm

Quantopian Lecture Series: ARCH, GARCH, and GMM
Quantopian Lecture Series: ARCH, GARCH, and GMM
  • www.quantopian.com
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by Quantopian. In addition, the material offers no opinion with respect...
 
Maxim Dmitrievsky:
here with

Here is an article and notebook from quantopian, maybe it is clearer there

I will hang out at that resource for a while, see what people do, maybe there is something interesting

https://www.quantopian.com/posts/quantopian-lecture-series-arch-garch-and-gmm


Looked at it, thanks!

Probably not bad for students of the relevant specialty.

If theory then it's primary source, if yes - literature on practical use of theory, if yes - code then only the one which can be used in the future for practical real life purposes.

So far rugarch meets all the criteria.

Nevertheless, thanks again, it's always nice to see something else.

 
SanSanych Fomenko:

I looked it up, thank you!

Probably not bad for students of the corresponding specialty.

I don't study new things that way: if theory, then primary source, we need literature on practical application of theory, if code, then only such, which can be used in the future for practical purposes on the real.

So far rugarch meets all the criteria.

Nevertheless, thanks again, it is always informative to look at something else.


Not at all :) of course you're right, if you study it deeply.

I've got a simple approach - to look through a heap of shit, choose the most interesting, then see if it has at least some trading potential and if it has then think how to use it with some experience and build a bot :) I'm not going to study stuff in depth unless I see it myself or someone inspires me that it's not a waste of time, because I have too much stuff for my eyes