Bayesian regression - Has anyone made an EA using this algorithm? - 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
Who knows what to do here?
You have to start somewhere, figure it out, scour the internet, collect information, maybe a source code or at least a sensible formula will turn up.
You don't have to dig around: "Everything was stolen before us"!
You take Rattle, there are 6 models, one of them is linear, you feed the data from Exel and at first you get excited, but not for long...
How to do it is described in my article"Random forests predict trends". For someone unfamiliar with R it takes an hour of work, and then you twist and turn...
On the wrong side of things. You have to take it, code it and check it out. You're already starting to... ...about the normality of the distribution.
You don't have to dig up anything - "it's all been stolen before us"!
You take Rattle, there are 6 models, in particular linear, you feed in data from Excel and get wildly happy at first, but not for long...
How to do it is described in my article"Random forests predict trends". For someone unfamiliar with R it takes an hour of work, and then you twist and turn...
This is not interesting (interesting of course, but not completely). To write an indicator for the terminal, an advisor. To understand the method itself.
It would be a good start if the formulas from the article could help translate into string form for mql.
If we discuss using models from Rattle in EAs, there are two steps:
1. the utterly fantastic: importing R code that implements models on MCLs. This is absolutely not realistic, and more importantly unnecessary.
2. the reference from MKL to R. It is realistic and not too much trouble: a few lines in MCL and a few lines in R itself which call the model itself. I went this way myself.
The problem is something else.
It's the models themselves. But the complete surprise for TA users is the data preparation (datd mining) and evaluation of simulation results. I am delighted with Rattle, which is essentially a GUI and does not require any preparatory work to get results in all three areas at once, with a few clicks of the mouse.
PS. Let's not forget that:
1. R for free is an extremely advanced system from the world's top in the field of statistics.
2. R is open source
3. R is extremely friendly in communicating with C.
About half a year ago I spent a few days (I was starting to dry up :)) to figure out how and what counts. Figured it all out (at least no questions left to me at the time, but now it's not all clear), but it didn't get to programming, and then forgotten.
Does the following lines confuse anyone
?
And the essence of the method
there is a fitting of coefficients over 4 parameters, three of which (parameters) are calculated based on "learning". This point - find 4 coefficients - is essentially a fitting and it seems to me that it will not work, while on the history you can draw anything by methods much easier. And the above article has become popular as no one can double-check the result. However, maybe I'm wrong1. utterly fantastic: importing R code that implements models on the MCL. This is absolutely not realistic, and more importantly not necessary.
...
About half a year ago I spent a few days (I was starting to dry up :)) to figure out how and what counts. Figured it all out (at least no questions left to me at the time, but now it's not all clear), but it didn't get to programming, and then forgotten.
Does the following lines confuse anyone
?
And the gist of the method
there is a fitting of coefficients over 4 parameters, three of which (parameters) are calculated based on "learning". This point - finding 4 coefficients - is essentially a fit and I don't think it will work, and you can draw anything on history with much simpler methods. And the above article has become popular as no one can double-check the result. However, maybe I'm wrongAboutk-means.
We have a bunch of indicators showing overbought/oversold zones.
I took RSI values for a certain interval and divided them into three classes.
The result.
1. The numbers indicated in the documentation are almost never 30/70
2. The 30/70 limits are variable and change as you move
3. An Expert Advisor that uses RSI with such variable bounds significantly improves performance, and most importantly, behaves more steadily.
Dmitry Fedoseev
Why don't you codek-means?
Or start to code base package with R? The documentation is ready. It's called "R reference". Over 2000 pages.
PS.
There are about 7000 packages in R containing over 130,000 functions. I think about 5% is relevant to us.
PSPS
R is a world top statistics and graphics system. Besides it there are 2 other comparable systems, but they are paid for.
...
1. How about codingk-means?
2. or will you start coding the basic package that R comes with? The documentation is ready. It's called "R reference". Over 2000 pages.
PS.
There are about 7000 packages in R containing over 130,000 functions. I think about 5% is relevant to us.
PSPS
R is a world top statistics and graphics system. There are 2 other comparable systems besides it, but they are paid for.
2. There was such a desire. Once I looked at how much stuff there is and how it's all connected. Do not.
1. Anything on the subject? Here on wikipedia:
The action of the algorithm is that it seeks to minimise the total quadratic deviation of cluster points from the centres of those clusters:
I would probably try 10 out of 10... //Bayesian doesn't seem to work at all.
http://datareview.info/article/10-tipov-regressii-kakoy-vyibrat/
What do you think?