Machine learning in trading: theory, models, practice and algo-trading - page 277
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
mytarmailS:
It depends, when a personis looking for something that works better ready-made packages, and when you have already found it, then you need to understand and have to rewrite your own code
don't you agree?
I agree with you on the whole.
The only problem is what to understand by the term "search", in what semantic space to search. Above you agreed with necessity of understanding low-level algorithm for its successful use, something similar is in search of "ideas", one thing is to read on forums or profile sites about trendy data analysis techniques, such as deep neural networks, forests, etc. It is another thing to find a package in R, to put a bunch of rows in it, and to pull knobs, meditating on grief, which would then "authoritatively" state "neural networks for forex do not work", and quite another thing to understand all this at the level of ints and dubs, to feel the texture of data, why a neural network or a forest does not work well, how to improve, etc. Another level of abstraction, you think not by someone voiced thoughts, but your own ideas, which even without clear verbalization, but already have a code and results :)
Now I'm trying to load data into Rattle, but I get an error:
Error in sort.list(y) : 'x' must be atomic for 'sort.list'
Have you called 'sort' on a list?
Now I'm trying to load data into Rattle, but I get an error:
Can't use double? Or do I have to prepare it correctly?By the way, the MYTH is approaching when my article will see the light, I'm sure many people will find it useful :-)
Tell me, do you write any kind of oanda story in any way?
There is an appendix in my article. Take a look at it as a sample.
I read your article, I think it will be instructive for a beginner reader to understand the basics of using MO in algorithmic trading. But you made a mistake in constructing the target,"To shift ZigZag indicator to the left " is not necessary, it shows the future as two-way filters (which take data from left and right of the current moment).
I read your article, I think it will be instructive for a beginner reader to understand the basics of using MO in algorithmic trading. But you made a mistake in constructing the target,"Shift the ZigZag indicator to the left " is not necessary, it already shows the future, as two-way filters (which take data from the left and right of the current moment).
No, it's not an error.
Rattle is a pretty sneaky joke.
On the one hand it allows you to see the whole cycle of machine learning: dataminig, modeling, model evaluation at a fraction of the cost.
On the other hand, it gives the illusion that models are easy to use. And this is not easy at all. Each element of the models used requires a very thoughtful attitude. And this is required at absolutely all stages of work on MI application. In particular, the choice of a target variable is not a simple problem either. It's possible to choose very tempting target variables and then it turns out that it's impossible to select predictors for them. If we talk about the ZZ in the article, strictly speaking, it is not clear at all WHAT is being predicted....
Success, and most importantly without illusions, there are no grails, it takes work and knowledge.
No, it's not a mistake.
rattle is a pretty mean-spirited joke.
On the one hand it allows you to see the whole cycle of machine learning: dataminig, modeling, model evaluation at a fraction of the cost.
On the other hand, it gives the illusion that models are easy to use. And this is not easy at all. Each element of the models used requires a very thoughtful attitude. And this is required at absolutely all stages of work on MI application. In particular,the choice of a target variable is not a simple problem either. It's possible to choose very tempting target variables and then it turns out that it's impossible to select predictors for them. If we talk about the ZZ in the article, strictly speaking, it is not clear at all WHAT is being predicted....
Success, and most importantly without illusions, there is no grail, it takes work and knowledge.
I completely agree with you. I'm writing an article and it is almost ready, the question of writing the codes in MKUL5 needs to be solved, but writing the current article inspired me to create the second work with the working title "Treatise on Inputs and Outputs". And it is in this work I want to highlight the complexity of the choice of the target function, because it is quite a philosophical question in the first place, and in the second - it is difficult to formalize. Well, we are waiting for the publication of the article "How Neural Networks Earn", then "Treatise on Input and Output Variables" and so on "Practical Use of Skynet when Working with a Sequence" Ie will be a series of three articles. Their titles are really working, except the first, although who knows, maybe I'll change it, the article has not been published yet :-) But I hope for support from the inhabitants of this branch, and most importantly a constructive criticism. Thank you!
I hope you have read all the articles that have already been published on the site on these topics and will not describe the invention of the bicycle?
New ideas are always interesting.
Good luck