Machine learning in trading: theory, models, practice and algo-trading - page 3151

New comment
 
mytarmailS #:

So how does AMO find a pattern from the example?

I described it all very modestly for the sake of clarity in reality it looks more like this.



And your model sees only this: (the last 5 unhourly candles).

Also note that there is no link to the index, if an important event was yesterday 200 candles ago, today the same event may be already 1555 candles ago or 12 for example....

AC will find such a pattern, AMO will not!

AMO needs each feature to always have the same column in the table, so that it is always triggered under the same index.


or like this, which is also pretty visual.

[[1]]
[1] ".....A...............................................................B........C..............SELL"

[[2]]
[1] ".......A........B...........C...........SELL"

[[3]]
[1] "........................................A........B.........................C.............SELL"

[[4]]
[1] "..A.............................................................B..............C...SELL"

[[5]]
[1] ".......A..........................................B...C...SELL"

and the modeller sees it.

[[1]]
[1] ".....A...............................................................B........C.............. SELL"

[[2]]
[1] ".......A........B...........C........... SELL"

[[3]]
[1] "........................................A........B.........................C............. SELL"

[[4]]
[1] "..A.............................................................B..............C... SELL"

[[5]]
[1] ".......A..........................................B...C... SELL"


Anyway, I hope I've made my point.

 
Aleksey Nikolayev #:

By the way, how's the python study going?

 
mytarmailS #:

By the way, how's the python study going?

It is a good language, but from some level it becomes too complicated for a non-programmer. For example, it is much harder to write extensions in C than in R. I really liked the tables in numpy.

 
Aleksey Nikolayev #:

I didn't dive into the rules. I have already written that I came to the application of formal grammars from the other side - I looked at the price as constructed by stochastic grammar. I gave up the approach precisely because of its cumbersomeness, which is bad first of all because it provokes overtraining.

Now I avoid heavy models. The main informal rule for me is that the heaviness of the model should correspond to the heaviness of the information in the sample.

Your model is heavy enough to be a full-fledged price model, but the actual sample of prices we have (even if we add other information) is not enough for such a model.

Naturally, IMHO

100%

 
mytarmailS #:

Also note that there is no link to the index, if an important event was yesterday 200 candles ago, today the same event may be already 1555 candles ago or 12 for example....

AC will find such a pattern, AMO will not!

AMO needs each feature to always have the same column in the table, so that it is always triggered under the same index.


or like this, which is also pretty visual.

Curious, doing exactly what I described about half a year ago.

I don't understand how your rules search for the value of a feature vertically without reference to the index - in my concept there should be a range of acceptable search - I don't understand your implementation.

 
Aleksey Vyazmikin #:

Curious, doing exactly what I described about half a year ago.

I don't understand how your rules search for the value of a feature vertically without reference to the index - in my concept there should be a range of acceptable search - I don't understand your implementation.

By usual algorithms of associative rules, different depending on the task.

I gave you a solution (code) to your problem then (half a year ago), have you forgotten?
 
Aleksey Nikolayev #:

It is a good language, but from some level it becomes too complicated for a non-programmer. For example, it is much harder to write extensions in C than in R. I really liked the tables in numpy.

holy question)

for market research - R or python?

 
mytarmailS #:

holivar question)

for market research - R or Python?

For market research by me, at the moment - R. I am not ready to vouch for others or myself in the future).

 
mytarmailS #:
The usual algorithms of associative rules, different depending on the problem.

I gave you a solution (code) to your problem then (half a year ago), have you forgotten?

I don't even know what code we are talking about - apparently something failed to run. What code are we talking about?

You claim that the depth is not important events in time - and how it is written by the rule - I did not understand.

 
Maxim Dmitrievsky #:

you can ask chatgpt for formula decoding if you don't understand any symbols.

Y|T = 1 test group outcomes (with tritment)

Y|T = 0 - control group (without)

Y - class label, Y0,Y1 - class labels without and with the tritment

T - tritment entered into the model (including predictor) or not entered (1;0)

E - expectation

Split at any point as you divide by test and train

If you don't do mixing, you get a biased estimate of ATE+bias

ATE is the average treatment effect of the exposure

sleepy, I may mix up the letters, but the logic should be clear.

I still don't understand the idea of dividing at any point in the sample. After all, the point, it seems to me, is to find just the point at which the effect of the factor changed. Maybe it is necessary to cycle through different parts of the sample and use genetics to find the one that was newly influenced by the predictor?

Earlier you accused others of not telling the truth, but you yourself do not make clear the meaning of these actions, as you see, for trading purposes.

I have not tested anything on this issue yet, as it is difficult to automate it in MQL5.

1...314431453146314731483149315031513152315331543155315631573158...3498
New comment
Reason: