Machine learning in trading: theory, models, practice and algo-trading - page 3042

 
Ask chatgpt, often answers more informative than a human :)

I want to do an experiment - write a bot from scratch only through prompts.
 
Maxim Dmitrievsky write a bot from scratch only through prompts.

Why? To outdo Lilith? Although what would ..... ... ... ... I'd like to)))))))))))))))))))))))) A really cool tool in skilful hands.)

 
Valeriy Yastremskiy #:

Why? To outdo Lilith? Although what would ..... ... ... ... sya)))))))) It's a really cool tool in the right hands.)

It's a meta level of programming, plus she knows python perfectly. I also do prompts by voice, lying on the couch. It often happens that I have ideas, but I'm too lazy to write this code again :).
 
Maxim Dmitrievsky #:
It's meta-level programming, plus she knows python perfectly. I also do prompts by voice while lying on the sofa. It often happens that I have ideas, but I'm too lazy to write this code again :)

I agree, the right question is a meta-level))))

Well and check the correctness of code execution)
 
Valeriy Yastremskiy #:

I agree, a properly posed question is a meta-level)))))

Well and check the correctness of code execution)
Give me a neural interface so I don't have to move my mouth.
 
I think it's a degradation
 
mytarmailS #:

Extracting a few "good" rules/strategies from the data...

Full step

1) data transformation and normalisation

2) model training

3) rule extraction

4) rule filtering

5) visualisation

ready code , just substitute your data



The question is that if you can find "working TCs" in random, what ways can you prove that the found TCs on real data are not random?

Alexey is doing it here, I wonder if there is any statistical test for this kind of tasks?

The main problem in applying matstat for such problems is that the search for TCs is carried out by selection from a large number of variants. It is always possible to choose something very beautiful from a large set of variants - by a simple example I once showed here that modelling prices as SB, you can always "find" a good hour of the week for trading. And that's only 120 variants to choose from.

The matstat does not say that the selected TS is necessarily bad, it only says that such a result CAN (NOT MUST) be just the result of selection from the SB.

 
mytarmailS #:

Extracting a few "good" rules/strategies from the data...

I get an error on startup

> sw <- embed(x = close,dimension = 10)[,10:1] # make slide window data
Error in h(simpleError(msg, call)) : 
  ошибка при оценке аргумента '.data' при выборе метода для функции 'embed': argument ".data" is missing, with no default
 

A purely theoretical question has arisen - can an ONNX model be used to derive another ONNX model. For example, the first model is used to periodically retrain on new data and update the working model. Meaning, without using python etc.

At first glance, this is unlikely to be possible, but in case someone has tried to do something like this.

I have not managed to get any meaningful answers from the AI - it writes that it can and cites references that have nothing to do with the question).

 
Aleksey Nikolayev #:

The main problem in using matstat for such tasks is that the search for TS is carried out by selection from a large number of variants. It is always possible to choose something very beautiful from a large set of variants - by a simple example I once showed here that by modelling prices as a CB, you can always "find" a good hour of the week for trading. And that's only 120 variants to choose from.

The matstat does not say that the selected TS is necessarily bad, it only says that such a result CAN (NOT MUST) be just the result of selection from the SB.

I still don't get it, there is no way to say with certainty whether the fin. res. is statistically significant or not? TC or not?

Aleksey Vyazmikin #:

I get an error when launching

1) Are the data the same as in the example?

2) Maybe in the new R the names of function arguments have changed

?embed