Machine learning in trading: theory, models, practice and algo-trading - page 1281
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
Try doing regression
trade only when the forecast has a high correlation with the price over N bars.
It's like a trend? When you understand that it's a trend, it's already coming to the end and it's time to exit rather than enter.
Regression (in the execution of the forest) will be tested, but later.
It's like a trend? When you understand that it's a trend, it's coming to the end and it's time to go out instead of in.
I don't know, they wrote it that way... metaphorically really, they thought it was so hard to understand.
I stand by my opinion: here reside two indisputable relatives of the venerable KsanKsanych (Fa). 1) Alyoshenka the son, who is overtaken by angry investors, and 2) grandson Kesha, who promises billions to everyone who reads his grandfather's creations.
Please do not confuse the two!
Please do not compare SanSanych with Alyosha, Wizard, etc. SanSanych is a pro, and Alyosha and Wizard are Forex-demotivators, who lost a couple hundred dollars and took offense at the whole world. I'm not grandson of SanSanych, I just respect him and find his article very useful, what can not be said about the articles and teachings of the Konyukh disciple, who collects links and terms to look scientific, but in fact is zero without a stick.
On the learning one, since the trees are untrained. In over-trained trees should have been on the test tree, since the tree would remember the noise as well.
I think it doesn't matter for the untrained ones.
But sample size is important. The larger it is, the more representative it is. And my training plot is three times larger.
---------
From the https://www.mql5.com/ru/blogs/post/723619 tutorial, a large representative sample makes balancing across classes unnecessary, reducing temporal randomness. Transferred this to the untrained trees.
But I may be wrong, and we need to check the significance of predictors on the test plot.
How do you determine if the tree is pre-trained or not (a forest of trees)?
I think, in any case, the methodology should consist of checking the stability of the tree on different samples (how this stability changes on different parts of sample/samples), training and testing. Otherwise you measure any noise, just lying well on history, i.e. you don't find regularities using the tree, you only describe the history of price movement and any predictors will do - the main thing is that they should be stable and frequent.
I'm looking for such patterns (leaves or mini catbust patterns), which are profitable, preferably every year (2014-2018) and meet some additional criteria.
Kesha grandson, let me look at the "trends" again, because all the jokers were banned, only you are left
How are your forests predicting trends?How do you determine if the tree is pre-trained or not (a forest of trees)?
I think, in any case, the methodology should consist of checking the stability of the tree on different samples (how this stability changes on different parts of sample/samples), training and testing. Otherwise you measure any noise, just lying well on history, i.e. you don't find regularities using the tree, you only describe the history of price movement and any predictors will do - the main thing is that they should be stable and frequent.
I'm looking for such regularities (leaves or mini catbust patterns), that are profitable preferably every year (2014-2018) and meet some additional criteria.
Stable and frequently occurring predictors associated with the target are patterns.
And stable and frequently encountered scientific terms in posts unrelated to the results are most likely overlearning:)
Stable and frequently occurring predictors associated with the targets and there are patterns.
Imagine the predictor is day/night, and you will have more than 1 in your targets during the day, for example, is this a good predictor? Or it's not about the day, but the fact that important news (affecting the market) is released more often during the day than at night.
And the stable and frequent sci-fi terms in posts unrelated to results is probably over-learning:)
I don't think it's appropriate to judge a person who does a lot to popularize MO here...
Kesha grandson, let me look at the "trends" again, because all the jokers were banned, only you are left
How are your forests predicting trends?I'm OK, stable 10-15% error rate, on tests. In real trading it's all mixed up and undefined, but I am a risk taker unlike you and similar students of studs who sit on the neck of elderly parents.
How do you determine if the tree is pre-trained or not (a forest of trees)?
I limited the tree to 1 line:
samples++; if(samples < 20){ then don't divide the node anymore, but leave the leaf}
i.e. there will be at least 20 samples left in the leaf, for representativeness.
That's the whole release you asked for)))
The degree of understudy, i.e. the number of examples in the sheet can be any 10, 100, 1000 or optimized. In xgboost it's called min_child_weight