Machine learning in trading: theory, models, practice and algo-trading - page 950
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
Why divide the file, if everything is already divided into two files? I just don't know how to do it in R, no one could explain it to me, I guess I'm stupid.
Maybe it's easier to use https://www.cs.waikato.ac.nz/ml/weka/downloading.html if you don't have time to study programming?
But here is a different model:
The result is QUANTITELY different, although the model is qualitatively different, should work poorly on your data.
We need to improve the randomForest
Got it, thanks, then I will deal with trees and forests - I like them big and ideologically.
Why divide the file, if everything is already divided into two files? I just do not know how to do it in R, no one could not explain to me, apparently stupid.
Dividing is a piece of cake, the problem is the prejudice against R.
I very much hope that the network will be able to outperform an optimized Expert Advisor on history :)
What is the network for?
No bias, just poor knowledge of the language, no Russian HELP (I already have one book, but the book should be read all through, unlike the HELP, and not sure what was needed there), in general, it is difficult to learn. And it's not clear why people don't like GUI so much, it saves time...
And about the network, I misspoke, it's just about the MO in general.
Where did you pick up so many farthers? Did you manually select the strategy? crazy :)
The logic of scaffolds should be the same.
I picked up these predictors from my bitter experience in manual trading, when I've lost money and I don't understand why I made a mistake when I entered the market. I have a problem - I do not like to lose money and that is why I have a hard time closing positions that cause me big troubles when I trade with hands. After such events you just work hard, testing, analyzing, looking for a solution to avoid a loss, generating ideas, testing them on the history, rejecting some of them, but not others. A lot of ideas are left without implementation because of the difficulty of programming them for me, but they remain on paper, papers fill the table...
Thanks for the reassuring answer about the scaffolding!
Maybe, it is easier to use https://www.cs.waikato.ac.nz/ml/weka/downloading.html, if it is boring to study programming?
Yes, I have this century - but I don't know how to use it!
And, then, how to make it work with MT5?
All this ***, in general, in Rattle in 2015 trained forest, default settings, produced this result
Summary of the Random Forest Model ================================== Number of observations used to build the model: 98573 Missing value imputation is active. Call: randomForest(formula = as.factor(arr_Buy) ~ ., data = crs$dataset[crs$sample, c(crs$input, crs$target)], ntree = 500, mtry = 8, importance = TRUE, replace = FALSE, na.action = randomForest::na.roughfix) Type of random forest: classification Number of trees: 500 No. of variables tried at each split: 8 OOB estimate of error rate: 6.89% Confusion matrix: -1 0 1 class.error -1 24452 2147 27 0.08164952 0 1138 42398 1180 0.05183827 1 36 2265 24930 0.08449928I learned means csv file to test the model on other data to load (for this file must first be opened as a file to work with the data, and then exported and already this exported file to open in the tab Evaluate) - loaded for 2016.
Got this dull result
What is this retraining, wrong settings, a radically different market?
Then why do I get better results on the tree in Deductor Studio with the same data?
Welcome to the world of curvafitting
By the way, I poked around EMD - decomposition has to be done on every new bar, because of this f-f is noisy because with the addition of new data all the mods jump back and forth. Nonsense, only suitable for single cases
nonsense nonsense nonsense nonsense nonsense nonsense. but I discovered a new way to manage positionsWelcome to the world of curvafitting
By the way, I poked around EMD - decomposition has to be done on every new bar, because of this f-f is noisy because with the addition of new data all the mods jump back and forth. Nonsense, only suitable for single cases.
I've learned a new way to manage positions.At first I thought, that's it, I've been scolded here, it turns out that the compound word has a different meaning...
Are you suggesting that the issue is that my exit from a position is not based on a pattern, but on a stop loss and this greatly distorts the result?
As for EMD, I have an idea to use this approach to create counter-trend channels...
What's the new way to manage positions?At first I thought, that's it, I've been scolded here, but it turns out that the compound word has a different meaning...
I was thinking that, at first I thought, well, that's it, I've been scolded here, but it turned out that the compound word has a meaning.
Regarding EMD, I had an idea to use this approach to create counter-trend channels...
And what is the way to manage positions?It is impossible to use EMD in dynamics because of the above reasons
It takes too much time to explain the method, everything is intertwined with RL
Yes and in your case - the result is expected on the new data. This is almost always the case. Partially solved by ensembles of independent modelsGot such a dull result
What is this retraining, wrong settings, fundamentally different market?
Then why am I getting better results on the tree in Deductor Studio with the same data?
The main proof of retraining: I did NOT find any noise predictors - all noise, that's why such good results in training.
EMD cannot be used in dynamics at all because of the reasons described above
about the way is long to explain, everything is intertwined with RL
And in your case - the result is expected on the new data. This is almost always the case. Partially solved by ensembles of independent modelsI'll add a couple more predictors and move on to ensembles.... and then I'll start dancing and tambourines.