Machine learning in trading: theory, models, practice and algo-trading - page 2725

[mytarmailS]

Evgeny Dyuka #:

There are methods that work.

What are they?

You can filter with filters, neuron or Mashka, the result is the same - lag.

[Maxim Dmitrievsky]

СанСаныч Фоменко #:

Very curious!

Maybe at least for a while the amount of rubbish, not even rubbish, but just nonsense on the thread will decrease.

I guess there are a few traps that lead to overtraining

The main one is that the features are irrelevant to the target (as you wrote before)

The second is outliers that bias the model

The third is a large number of stationary but uninformative features. Overfitting is obtained due to the difference of features that are not relevant to the target

[Maxim Kuznetsov]

On a more personal note, has anyone got any results anywhere? Anything at all.

So far, I've heard one man say that his good friends know it's promising.

Here it is (DL NN) is still at this level. All attempts to drive profits out of abstract time series are still 50/50.

of course, there is a variant that who found it has gone to the land of eternal spring and swarthy maidens on a private yacht, and who had a bummer that embarrassedly keeps silent...but all others in terms of efficiency do not go anywhere from other methods.

[mytarmailS]

Maxim Kuznetsov #:

So far, I've heard one man say that his good friends know it's a promising business.

))))))) genius.

[Maxim Dmitrievsky]

Maxim Kuznetsov #:

On a more personal note, does anyone have any results anywhere? Anything at all.

So far, I've heard one man say that his good friends know it's a promising business.

Here it is (DL NN) is still at this level. All attempts to drive profits out of abstract time series are still 50/50.

Of course, there is a variant that who found it has gone to the land of eternal spring and swarthy maidens on a private yacht, and who has a bummer that embarrassedly silent...but all others in terms of efficiency do not go anywhere from other methods.

If you ask 3.5 people in this thread, then 2.5 do not even have realisations in the form of bots to be able to test something and draw conclusions, half of them can not program, the second has never opened a terminal. And the remaining ones are not a representative sample

And the average level of development is, to put it mildly, not geniuses. We should probably turn to some other statistics.

[Aleksey Vyazmikin]

Maxim Kuznetsov #:

On a more personal note, does anyone have any results anywhere? Anything at all.

So far the results are the same as in other methods - you can create 100 models and 50 of them will work on completely new data, but how to determine which ones will work is a mystery.

Perhaps the solution is only in batch methods, in creating models that are not similar to each other for diversification.

[Evgeny Dyuka]

I'm distracting from an interesting discussion, I have a practical question

int file_handle=FileOpen(fileName,FILE_READ|FILE_TXT|FILE_ANSI);

I access a file to read it, but how do I know that it is currently available for reading?
if it is unavailable, what happens?
the help doesn't say anything clear about it.

[Maxim Kuznetsov]

Evgeny Dyuka #:

to divert from an interesting discussion, I have a practical question.

I access a file to read it, but how do I know that it is currently available for reading?
if it is unavailable, what happens?
the help doesn't say anything clear about it.

if you can't open it for reading, INVALID_HANDLE will be returned and you can find out the cause of the error via GetLastError().

sometimes you can ask FileIsExists in advance just in case - to check if there is such a file at all.

[Aleksey Vyazmikin]

Evgeny Dyuka #:

to divert from an interesting discussion, I have a practical question.

I access a file to read it, but how do I know that it is currently available for reading?
if it is unavailable, what happens?
the help doesn't say anything clear about it.

The help says that there will be an error, there is a code example in the help

//--- правильный способ работы в "файловой песочнице"
   ResetLastError();
   filehandle=FileOpen("fractals.csv",FILE_WRITE|FILE_CSV);
   if(filehandle!=INVALID_HANDLE)
     {
      FileWrite(filehandle,TimeCurrent(),Symbol(), EnumToString(_Period));
      FileClose(filehandle);
      Print("FileOpen OK");
     }
   else Print("Операция FileOpen неудачна, ошибка ",GetLastError());

[Maxim Dmitrievsky]

You can also check an already open file

https://www.mql5.com/ru/docs/constants/io_constants/enum_file_property_integer

FILE_IS_READABLE