Machine learning in trading: theory, models, practice and algo-trading - page 859
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
I would like to add more on reinforcementlearning. There are two packages on this topic in R: ReinforcementLearning and reinforcementlearning. The first one does not use Pyhton modules and implements simple algorithms. The second uses Python::gym and implements a large variety accordingly. The latest in Python are Tensorforce and keras-rl (based on tensorflow/keras)/ Unfortunately there is almost zero documentation. But there are a lot of enthusiasts who share their developments and examples.
Haven't got around to it yet.
Try
)))
For the first search, it is advised to use either "peng" (faster) or "esteves"
(more reliable but much slower for large datasets) and, in case the number of
variables is large (>100), restrict the "forward" search to "n.var = 100." The
progress bar will give you an idea of the remaining run time.
library(varrank)
data(nassCDS, package = "DAAG")
nassCDS.varrank <- varrank(data.df = nassCDS,
method = "peng",
variable.important = "dead",
variable.method = "dead", variable.method = "sturges",
algorithm = "forward",
scheme = "mid",
verbose = FALSE)
summary(nassCDS.varrank)
plot(nassCDS.varrank, notecex = 0.5)
Originally tried with "esteves" it ate all the memory even at 500 rows. At 100 rows it used about 5 gigs.
With peng was able to take all 13x6400 data without eating up all the memory:
Ordered variables (decreasing importance):
5 6 8 12 74 1 2 11 10 3 9
Scores 0.104 -0.005 -0.021 -0.03 -0.101 -0.114 -0.139 -0.128 -0.181 -0.173 -0.227 NA
But the sort order is very different from the other packets. And it is not clear how it sorts, Scores-not sorted.
The average of the 37 packages from CoreLearn:
5 1 7 11 9 4 3 6 10 2 8 12
Average of 37 packages from CoreLearn:
5 1 7 11 9 4 3 6 10 2 8 12
There are two functions in CORELearn: attrEval(), ordEval()
Do you use both?
PS
Your progress in R is fascinating!
There are two functions in CORELearn: attrEval(), ordEval()
Do you use both?
PS
Your progress in R is amazing!
ordEval - I did not understand how to use, I did not see any sorted predictors, or weights by which I could sort myself. I think attrEval is enough with 37 algorithms. In fact, I got an ensemble of methods, by which I determine the importance of predictors.
If you yourself are not able to solve such problems, find me a module VisSim NeuralNet, I will show you how to do it.
This particular algorithm, does it select predictors good or bad?
In general, in the selection of predictors, what is good and what is bad?
The little son came to his father, and the little son asked
Bullshit. And R with MO is bullshit. Eviews is the most traveled program.
That's not a fact. Here is another program on the principle of on/off dough. But only for those who use cycles. Exclusively the work on laying out the quotes in the spectrum of cycles. There are quite a lot of settings. There are networks and so on.
http://www.timingsolution.ru/https://radikal.ru/video/qD9j3mOipVf
wow wow wow wow wow wow wow wow wow wow wow wow wow wow wow wow wow wow wow wow wow wow wow, you've mastered the video editor just for me???? Did you know you can use it for free. In fact, I think the program is okay, but taking into account the use of R and the knowledge that I now have. More precisely, the vision of the market and the field of MO in general!!!!! But I don't want to study it. Already thought of everything. I just want to work according to the algorithm. One, the only one and without any deviations and search there....
That's great! So why aren't you trading then? All these pictures are worthless without any real trading. What's the problem?
wow wow wow wow wow wow wow wow wow wow wow wow wow wow wow wow wow wow wow wow wow wow wow, you've mastered the video editor just for me???? Did you know you can use it for free. In fact, I think the program is okay, but taking into account the use of R and the knowledge that I now have. More precisely, the vision of the market and the field of MO in general!!!!! But I don't want to study it. Already thought of everything. I just want to work according to the algorithm. One, the only one and without any deviations and search there....
Heh))) Target on the horizon I approve.))
Like come to the bank: So, like, how much today?
- The balance of your trading account has been replenished today by 5,508.56 euros.
- Well, let's make it to tomorrow somehow...
- Would you like to take out a discounted insurance policy?
- No, thanks...