Machine learning in trading: theory, models, practice and algo-trading - page 2193

 
mytarmailS:

So here's the deal

I want to predict channel beams

...

Karoch should solve the problem, who knows how to train networks with three outputs (parameters) on the regression?

Or have other suggestions ...?


PS. three models to train is not an option ...

So give people the data, collect and present it in a convenient format, csv, for example.

 
mytarmailS:

So here's the deal.


Only three values...


Or are there more suggestions... ?


No change, ranges of (reasonable) values, step, and combinatorics. There are no other logical solutions. Except hash sums)

 
Aliaksandr Hryshyn:

So give people the data, collect it and present it in a convenient format, csv, for example.

Alexander, the data will not help here, I am interested in whether there is a methodology to predict n values at once...

If you train AMO, and even successfully, it won't make me feel better, unless you program in R...

Valeriy Yastremskiy:

No change, ranges of (reasonable) values, step, and combinatorics. There are no other logical solutions. Except hash sums)

OK

 
mytarmailS:

Alexander, the data won't help here, I'm interested if there is a methodology to predict n values at once...

If you train AMO, and even successfully, it won't make me feel better, unless you program in R...

OK

Or if you watch lingvonNS, there are words. How about encoding 3 values into one word?

 
mytarmailS:

Anyway, here's the deal

the question is removed, I think I found solutions

https://www.kaggle.com/general/20017

 
mytarmailS:

Alexander, the data won't help here, I'm interested if there is a methodology to predict n values at once...

If you train AMO, and even successfully, it won't make me feel better, unless you program in R...

OK

I have everything. Here I have STATISTICA in my program, in Pyton there is definitely a solution, just experienced in this business could help you with the choice of approach, you need the data for research, to test the idea...

 
Aliaksandr Hryshyn:

Everything is there. I have STATISTICA in my program, in Pyton there is definitely a solution, just experienced in this business could help you with the choice of approach, you need the data for research, to test the idea...

ok,

I'll train the model, I'll post the data...

I accidentally deleted the script that created the dataset.



I know beforehand that there will be a huge error, I will probably have to make some forecasts.

 
The channel is built based on three extrema, there are only two options, or two extrema at the bottom and one at the top, or vice versa. It cannot be plotted on a two-dimensional chart in any other way, no matter how twisted you want it to be. It works 100% of a hundred times at any dimensionality. It is unclear what you are going to teach and train there and why.
 
Aliaksandr Hryshyn:

So give people the data, collect it and present it in a convenient format, csv, for example.

here , "up" "dw" "fast" for the forecast , the rest are signs

Files:
dd.zip  3211 kb
 
Wizard2018:
It works 100% of the time, at any size. What are you going to teach and train there, and why, I do not understand.

Where do you come from?