Discussion of article "Random Forests Predict Trends" - page 2
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
Of course it would be right to tell everything in order about R language, machine learning models on concrete examples and with real implementations in conjunction with MCL. It is not clear how to organise this. Just a thread, a series of articles with discussions? I'm not sure that the administration is interested. I will prepare a proposal. Let's see how they react.
It is more important whether the reading public is interested in this topic.
There are mostly programmers here and they are deeply immersed in the depths of C. They are nice and cosy there.
I'll think about it.
Of course it would be right to tell everything in order about R language, machine learning models on concrete examples and with real implementations in conjunction with MCL. It is not clear how to organise this. Just a thread, a series of articles with discussions? I'm not sure that the administration is interested. I will prepare a proposal. Let's see how they react.
It is more important whether the reading public is interested in this topic.
There are mostly programmers here and they are deeply immersed in the depths of C. They are nice and cosy there.
I'll think about it.
It's just a matter of course. We open a thread and that's all. On the fourth forum I have several threads on econometrics. I just got bored.
And there is a base here. The forum is full of people working with neural net. Material that would show that NS is not a godsend, I think would hook this audience. Especially if they realise that it is possible to raise the profit factor a couple of times by changing the model. And these people are very prepared.
If we're talking about education in Russian, I have two translations of the documentation, plus I wrote a book (390 p) on Rattle and the ideology of the models used in it.
It's just an open-and-shut thing. We open a thread and that's it. I have several econometrics threads on the fourth forum. I just got bored.
And the base is here. The forum is full of people working with neural net. Material that would show that NS is not a godsend, I think would hook this audience. Especially if they realise that it is possible to raise the profit factor a couple of times by changing the model. And these people are very prepared.
If we're talking about education in Russian, I have two translations of the documentation, plus I wrote a book (390 p) on Rattle and the ideology of the models used in it.
I'll finish the article this weekend and send it for review. After that we'll see.
What topics do you suggest we look at? In what order? Have you thought about it? If we start, we should start with a description of the language At least a fluent one.
Writing takes a lot of time. There must be motivation.
Let's think about it together. We'll have to pull in some people from the last thread.
Good luck
I'll finish the article this weekend, send it out for review. After that, we'll see.
What topics do you suggest we look at? In what order? Have you thought about it? If we start, we should start with a description of the language At least a fluent one.
Writing takes up a lot of time. You have to be motivated.
Let's think about it together. We'll have to pull in some people from the last thread.
Good luck.
1. Probably from the language, but: to take the minimum and if possible only analogues from MKL, so that people can see - there are no difficulties, even more convenient because of the interpreter. For example, narrow down the concept of "object" to MKL, replace matrix operations with loops if possible - in general simplify.
2. For me the motivation is to search for new ideas. From the last forum from econmodel I got the problem that occurs on consecutive samples. I managed to repeat it and solve it. It turned out to be a question of principle. In general, we need to get together.
PS. The number of forumers who have written off the interface with R over 1000! True, earlier all of them were on the four. So the question on what forum to open.
Error 0.15, that means 85% chance of guessing the trend! Scaffolding itself is not an end in itself, our goal is to make money. What did the trade show? Is there a balance chart?
The main problem of predictive models is the selection of initial data. Rattle is a very convenient tool for solving this problem and the article is aimed at this.
I am willing to discuss the raw data with getting an estimate. If someone provides a source csv file, I am ready to do calculations and post the result here.
Anything else is beyond the scope of the article.
How can I see the full set of independent and dependent variable values? Can it be posted in a separate file in tabular or text form?
How can I see the full set of independent and dependent variable values? Can't you put it in a separate file in tabular or text form?
I think so (attachment), although I've already forgotten the article.
Why don't you try it yourself? The attached file is deliberately redundant so that the reader can not only repeat the calculations in the article, but also check his ideas.
I think so (attachment), although I've already forgotten the article.
Why don't you try it yourself? The one attached to the article is deliberately redundant so that the reader can not only repeat the calculations in the article, but also check his ideas.
Probably because I "tried it myself" and the results are different and not for the better? Makes sense, doesn't it?
No, it is the original set of all variables that is needed - perhaps I don't have the same independent variables
Probably because I "tried it myself" and the results are different and not for the better? Makes sense, doesn't it?
No, it is the original set of all variables that is needed - perhaps I don't have the same independent variables
In the attachment above, the results, and I emphasise very similar to those in the article. Why "similar"? All random forest algorithms involve a random number sensor (this is considered a virtue of the algorithm), so the results may be similar, but may not differ dramatically.
It would be interesting to see your result.
If about the estimation of the znA value of variables in the random forest algorithm. I have not managed to use it (though it is not an indicator - just experience). There are other, more constructive algorithms, if you follow them, you can seriously improve the result.
In general, the above article is a window into the world of classification models. It's just a quick try and evaluate. And then tedious and dull work of selecting predictors.