Machine learning in trading: theory, models, practice and algo-trading - page 907

 
Oh, man... I wanted to do a TS on Boshkas, and without stochastics, it fails. I've been bored, it's been an hour. I fixed the bugs, but it's not drawing anything...... It's a mess, fellas. Wait for the continuation of the next series as they say :-(
 
Mihail Marchukajtes:
What a pisser. I wanted to make a TS for Boshki, but it is a failure without stochastics. I have exhausted myself for more than an hour. I fixed the bugs, but it does not draw anything...... It's a mess, fellas. Wait for the continuation of the next series as they say :-(

An hour is not enough....

2-3 days

Keep working, don't get distracted.

 
Maxim Dmitrievsky:

I gave advice about google, I heard about amazon's, but I do not think it's so bad.

google is on fire, you can choose the location you want, 100% stability, 100% protection

Thanks, with google figured it out. For registration, one more visa ordered myself (required for registration, although 1 year for free), so as not to shine the main. In 3-5 days it will come, and then I will continue the epopee).

 

What software has the ability to enumerate predictors from the set?

Because it turned out that the tree cuts half of the predictors, and individually they are interesting, but it is tedious to try all the combinations by hand....

 
Aleksey Vyazmikin:

What software has the ability to enumerate predictors from the set?

Because it turned out that the tree cuts half of the predictors, and individually they are interesting, but it is tedious to search all combinations by hand....

Vtreat.R

 
Yuriy Asaulenko:

Thank you, I figured it out with Google. I ordered myself another Visa for registration (required for registration, although 1 year free), so as not to shine the main one. In 3-5 days will come, then continue the epic).

Yes, no credit. The money will not be written off at the end of savings, just a notice will come that ran out

 
Aleksey Vyazmikin:

What software has the ability to enumerate predictors from the set?

Because it turned out that the tree cuts half of the predictors, and individually they are interesting, but it is tedious to search all the combinations by hand....

In cycles))

But all combinations and PCs will take twenty-four hours to recalculate...
10 predictors, combinations 2^10 = 1024

20 predictors, = 2^20 = 1048576

That's how many times you have to train a les/nS.

 
Doc and I were discussing the results of elmnn model training the other day. The script is attached to his blog. So he was very pleased with the quality of the models I received. Hence the Conclusion: Forests, MLPs, Deep Learning all pale before a set of good predictors. You need to prepare them properly and then you can use ANY of the models suitable for the task and the elmnn network is proof of that!!!!!!
 
Mihail Marchukajtes:
I've been discussing the results of elmnn model training with Doc the other day. The script is attached to his blog. So he was very pleased with the quality of my models. Hence the Conclusion: Forests, MLPs, Deep Learning all pale before a set of good predictors. You need to prepare them properly and then you can use ANY of the models suitable for the task and the elmnn network is proof of that!!!!!!

Will there be any pictures? Stop talking about nothing, we don't need any good ones either, lol :)

 
Mihail Marchukajtes:
I've been discussing with Doc about the results of elmnn model training. The script is attached to his blog. So he was very pleased with the quality of my models. Hence the Conclusion: Forests, MLPs, Deep Learning all pale before a set of good predictors. You need to prepare them properly and then you can use ANY of the models suitable for the task and the elmnn network is proof of that!!!!!!

+100!!!


Found a set of predictors, even a simple rpart gives 3-4% worse results than more advanced models, and so all models can be used in the TC. No retraining at all, from the word TOTALLY