Machine learning in trading: theory, models, practice and algo-trading - page 1155

New comment
 
mytarmailS:

Windows or linux do you have?

Windows 7 64x

 
Aleksey Vyazmikin:

Windows 7 64x

I'm not sure but I think you downloaded the package for linux


 
mytarmailS:

I'm not sure but I think you downloaded a package for linux


This is the command you gave.

I tried to zip it after unzipping, but it did not work (no error message either). Where can I get the right file for Windows?

 
Aleksey Vyazmikin:

This is the command you gave.

I tried to zip it after unzipping - did not work (although no error messages). Where then download the correct file for Windows?

I do not know what the problem is.

Ask your question on stackowerflow or write to these *** of Yandex

 
mytarmailS:

I don't know what the problem is anymore.

Ask a question on stackowerflow or write these *** of Yandex

Thanks for trying to help.

Tell us about your results of using this package, please.

 
Aleksey Vyazmikin:

Thank you for trying to help.

Tell us about your results of using this package, please.

I didn't use it, it's the predictors that matter, not the model, the model may perform 0.5%-3% better than another one, while the signs determine everything

 
mytarmailS:

I did not apply it, it is the predictor signs that are important, not the model, the model can work better than another by 0.5%-3%, and the signs determine everything

Well, I wouldn't say so, different approaches give different results, for example on my data some neuronics work, others do not, but the tree works for example and Kohonen maps and some other clustering ...

 
By the way, can anyone know a good NS or another method, for conventionally matrix images, like pixels? I have in the form of numbers from 1 to 20 matrix values, i.e. marked points on the matrix 20 but 20.
 
mytarmailS:

I did not apply it, it is the predictor signs that are important, not the model, the model can work better than the other by 0.5%-3%, and the signs determine everything.

It is necessary to bluntly do preprocessing of initial BP of returns, smooth outliers, etc. in order to obtain stationarity.

According to Kolmogorov, only returns and cumulative sum of returns in a sliding window may be predicted, as they have expectation=0. But not the price itself and other stuff.

And how do price and returnees relate? Right - price is an integral of all returns.

 
Aleksey Vyazmikin:
By the way, who knows a good NS or another method for conditionally matrix images like pixels? I have a matrix value from 1 to 20 in the form of numbers, i.e. marked points on the matrix 20 but 20.

Google it, there is a lot of stuff in the same R-ka

Alexander_K2:

It is necessary to bluntly pre-process the initial BP of returns, smooth outliers, etc. in order to obtain stationarity.

According to Kolmogorov, only returns and cumulative sum of returns in a sliding window may be predicted, as they have expectation=0. But not the price itself and other stuff.

And how do price and returnees relate? Right - the price is the integral of all returns.

Yes, yes, we've heard it 100 times ... and tried it 100 years ago and threw it away because it's garbage

1...114811491150115111521153115411551156115711581159116011611162...3659
New comment