Machine learning in trading: theory, models, practice and algo-trading - page 23

 
mytarmailS:

How do I make the code not to make extra columns? For example, 3 columns in the function produces 9 combinations, although in fact it is enough to have three, like in my example above

there is no sense to do A/B and then B/A

это думайте сами :) ваш R
 
Alexey Burnakov:
это думайте сами :) ваш R
Ahaha... Thanks to you! )
 
mytarmailS:
Ahaha... Thanks for that too! )

seriously. no one can learn how to program except you.

ok, i'll give you a hint this time:

if(j > i) { ....... }

that's it, guys )

 
Alexey Burnakov:

seriously. no one can learn how to program except you.

ok, i'll give you a hint this time:

if(j > i) { ....... }

that's all, guys )

Your code died on 30 iterations))) lack of memory, what else can you do? I spent so much time on writing all the features I had only to mix in to check the model, and then this is bullshit
 
mytarmailS:
Your code died at 30 iterations))) not enough memory, what else can you do? spent so much time writing all the features left to mix in the model check, and then this crap

I've never heard that one before! "your code is dead". It's R that died, but the code is alive and well and on the data I cited it works for less than a second.

Everything works for me. As a variant, you really don't have enough memory and very big data arrays.

In general, this algorithm uses the basic functions and features of the language. It's hard to make it simpler and more compact if your R already crashes on some 30th iteration. The lack of memory means that in order to deal with leerning machines you will have to splurge a bit on a couple of memory modules. It's a harsh reality )

 
Alexey Burnakov:

I've never heard that one before! "your code is dead". It's R that died, while the code is alive and well and works for less than a second on the data I cited.

Yes, I agree, I was wrong... Then I made it a little easier, found a matching combinations package written in cp and tried it, same story there... My computer is really useless, and the combinations turned out 2500++ for 100 000 lines, R is just dying... I need to buy more memory unequivocally
 
mytarmailS:
Yes I agree, that's not what I said... Then I made it a little easier, found the appropriate combinations package written in cp and tried with it, it's the same story... My computer is really useless and I got 2500++ combinations per 100 000 lines, R is just dying... I need to buy more memory for sure
I think there is a variant of R together with SQL. There are no restrictions on array size.
 
Hi all. I too am fond of machine learning. Tried to use naive Bayesian classifier and support vector method so far.
 
SciFi:
Hi all. I'm into machine learning too. Tried so far to apply naive Bayesian classifier and reference vector method.
Shove everything into rattle and post the results, which should prove that your models are not overtrained.
 
mytarmailS:
Yes I agree, I didn't say it that way... My first time I tried it I found my combinations package written in cp and tried it and it was the same story... My computer is really useless and the combinations turned out to be 2500++ for 100 000 lines, R is just dying... I need to buy more memory unequivocally

It's about the physical limitations of your machine, unambiguously.

And in general, questions about the code should first try to solve themselves. Open google, type, for example, data frame columns in cycle results list stackoverflow.

And the site mentioned above has answers to 99% of questions.