Machine learning in trading: theory, models, practice and algo-trading - page 2647
![MQL5 - Language of trade strategies built-in the MetaTrader 5 client terminal](https://c.mql5.com/i/registerlandings/logo-2.png)
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
wai !
I understand you, you have special education and experience, unlike "activists".... (you have economics and professional experience in divorce centres?).
but how many times can you make fun of them?
Oh, the Stirlitzes/counterstirlitzes are here.)
It's all right... normal channel with bends and foreign inclusions, but the channel is stable)))).
It seems quite normal to me that a simple algorithm only works on a limited portion of the set of predictors.
You could do a step-by-step removal from the sample of what falls into the "good" boxes and apply the algorithm to the remainder. In essence, it would be similar to bousting. Probably, you can do it similarly to randomforest - take a large set of predictors and for each subset find several boxes.
If I understand correctly, the point of combining predictors is to arrange the identified boxes in such a way that they do not overlap each other, and if they do, they do not significantly worsen the result, right?
I'm just messing with you. I'm telling it like it is.
You just haven't seen the behind-the-scenes.
;)
you just haven't seen the behind-the-scenes.
;)
If I understand correctly, the point of combining predictors is to arrange the identified boxes in such a way that they do not overlap each other, and if they do, they do not significantly worsen the result, right?
I think the question of what to do with the detected boxes is complex and hardly has clear unambiguous rules for all possible cases. A good, well-thought-out algorithm is probably quite a secret "know-how")
If the cases are obtained on the same set of predictors, their non-intersection is probably enough. If there is an intersection, it can be allocated to a separate box, and its complements can be split into several boxes. However, too large a number of boxes will fragment the sample too much. Therefore, we can generalise the notion of a box - in the language of rules, this means adding negations and OR to AND.
If the boxes are obtained on completely different predictors (for example, by the randomforest method), then they can overlap only in the sense of the parts of the sample that fall into them. Some near-portfolio ideas are probably needed here.
If the predictor sets overlap partially, then probably some mixture of approaches, it is hard to say for sure.
It is not clear to me how this can be put into a unified scheme. The standard way of constructing decision trees simply and "nicely" circumvents these problems, which makes it not quite suitable for our purposes. It may be possible to improve it by selecting a pruning algorithm, but in my opinion it is better to creatively rework the rule construction algorithm.
I've seen it. We were laughing at your monkey not so long ago.
It makes sense, even a puppet theatre has a backstage.)
Makes sense, even puppet theatre has a backstage)
:D
I think that the question of what to do with the identified boxes is complex and hardly has clear, unambiguous rules for all possible cases. A good, well-thought-out algorithm is probably quite a secret "know-how").
If the cases are obtained on the same set of predictors, their non-intersection is probably enough. If there is an intersection, it can be allocated to a separate box, and its complements can be split into several boxes. However, too large a number of boxes will fragment the sample too much. Therefore, we can generalise the notion of a box - in the language of rules, this means adding negations and OR to AND.
If the boxes are obtained on completely different predictors (for example, by the randomforest method), then they can overlap only in the sense of the parts of the sample that fall into them. Some near-portfolio ideas are probably needed here.
If the sets of predictors overlap partially, then there must be some mixture of approaches, it is hard to say for sure.
It is not clear to me how this can be put into a unified scheme. The standard way of constructing decision trees simply and "nicely" circumvents these problems, which makes it not quite suitable for our purposes. It may be possible to improve it by selecting a pruning algorithm, but in my opinion it is better to creatively rework the rule construction algorithm.
Smartest conclusions...
Out-of-the-box algorithms have only approximating rules, i.e. dumb as hell, like:
X[1]>0.5 && X[3]> -0.2 &...
They have only variables (columns of matrix or frame X1.....X10) and abstract boundaries (digits X[1]>0.5 ), which he himself invents in the process of learning to create boundaries in classes.
What about the trivial X1 > X2
or the negation of X1 !> X2.
What about X1> (X2*X3)?
And binding to indices, which doesn't work on non-stationary market?
I mean the expression: if there was "A" but no "B" then "C".
It's just not for the algorithms out of the box.
The algorithm is the same forrest just creates hundreds of rules and looks at their sum of class probability. In the market we care about rare events, so we need to move to the number of rules, not the number of rules.