Machine learning in trading: theory, models, practice and algo-trading - page 3461

[[Deleted]]

fxsaber #:
Does the price (row) contain time data?

Yes

It is as if you are completely unfamiliar with econometrics - with the basics of time series analysis

All that many people write is some kind of self-consciousness, so they don't understand each other.

[mytarmailS]

Maxim Dmitrievsky #:
Yes

It's as if you are completely unfamiliar with econometrics - the basics of time series analysis

All that many people write is some kind of rubbish, so they don't understand each other.

add a pinch of pseudo-secrecy and pseudo-importance to each writer and you will realise that there is nothing and no one to communicate here )

[fxsaber]

Maxim Dmitrievsky #:
Yes

How to get time from a price series?

[[Deleted]]

fxsaber #:
How do you get time from a price range?

That's it, I'm drinking beer.

[[Deleted]]

mytarmailS #:

add a pinch of pseudo-secrecy and pseudo-importance to each writer and you will realise that there is nothing and no one to communicate with )

Conspiracy is the most hilarious :)

If they want to teach Teachers something, let them send money to the card at once.

[[Deleted]]

Ivan Butko #:
If we input price increments with some transformation (0...1) or (-1...1), the system will work infinitely.

Infinitely bad.

If you give pure prices as input and predict the next price as output, a simple MLP produces the following picture:

The best quality averaging that can be had.

That said, it only works on the forward if the price chart does not change dramatically. At some point the NS simply stops working.

And even additional training does not help, it just fades on new data and does not open. Then, when the market either calms down or something else happens to it, the NS after training with direct prices continues to work.

And sometimes it works very well. Especially on EURGBP and other flat markets.

UPD

Stationarity seems to kill some... important information

Well not infinitely bad, but worse, yes. You can do a transformation that is worse than the original series, but better than the retournals. That was the story.

One way was proposed back in the days of arima - fractional differentiation. The so-called FARIMA. But it is little better than returns.

[Renat Akhtyamov]

fxsaber #:
How do you get time from a price series?

as usual

go to the element of interest and ask him for price and time.

and the other way round, yes, it's the same as asking:

why don't stars fall ?

[fxsaber]

Renat Akhtyamov #:

I don't get it. There is some volatility with a period of 20, which correlates with the time of day. If you ask, it's bullshit.

[Renat Akhtyamov]

fxsaber #:

I don't get it. There is some volatility with a period of 20, which correlates with the time of day. If you ask, it's bullshit.

20 minutes?

[Aleksei Kuznetsov]

fxsaber #:
How to get time from the price series?

I think it's easier to add some columns for MO: day of week, hour, minute. Maybe also month, day of month and/or year.