Machine learning in trading: theory, models, practice and algo-trading - page 1357
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
On validation comes to a halt, which means that if further sampling is more like validation than training, we will not train something that will not be repeated. Another thing is that there are often intermediate junk trees between the improvement of validation readings and the training itself - we should weed them out...
The main question here is whether the past is repeating soon, how the present is changing, at what rate or jumps - the answers to these questions would give a lot of information on how best to build a sample.
Well, the main problem is the lack of data, I have less than 10k rows to train and validate.
That's why I raised the question about the proper scaling of quotes, if you arrange them by levels then they will be like quantum transitions.
there was a book somewhere about quantum mechanics in finance
That is why I raised the question about the correct scaling of quotations, if you arrange them by levels then there will be some kind of quantum transitions
There was a book somewhere about quantum mechanics in finance.
And you, Brutus, are a quantum mechanic. (What a contagion.)).
And you, Brutus, are a mech quantum. What a contagion.)
No, I just stumbled across the book by accident.
It's very little, especially if you don't know what you're looking for. And even if you do know, it's not a fact that it's there.
General words, of course, but you have to look for something that is everywhere and always and regularly repeated. Otherwise, you can't teach ME.
Yes, not enough, but what to do... It is necessary to invent another strategy, what would be a lot. But the set of predictors that haven't been analyzed before works for some symbols, like Sberbank.
That is why I raised the question about the correct scaling of quotations, if you arrange them by levels then there will be some kind of quantum transitions
There was a book somewhere about quantum mechanics in finance.
So I don't use quotes in their bare form. I use ATR scaling that allows to save form of information when volatility changes, but the information about volatility in absolute terms can also be useful.
I have another problem - there are already a lot of predictors that should be transformed - earlier there was an idea to replace them with ready generalizing rules - stable leaves. Experiments that year showed good results, but all work on splitting the tree went to waste - I've already said that's why I've started from the beginning, and it will take a very long time.
Yes, not enough, but what to do... We have to invent another strategy to have a lot. But a set of predictors works on some instruments that have not been previously investigated, the same Sber.
I'm working on 1m. In 3 months 55k OHLSV. What kind of history would I need if I switched to 1h, or at least 15m? And so I have 3 months, well 6 months. for everything - about everything.
Further, for 5-10 m unlikely that something drastic will happen, and even will - we react. And in an hour, very likely there will be an event, which may not be detected at once, but in the aftermath may disrupt the "planned" course of events. Well, and there is no pattern that you were beginning to count on - dissolved.
In general, my concept - to predict anything at long intervals is unrealistic. Classification, especially a priori, is the same prediction. And the same type of events will not be enough for classification.
I work for 1m. In 3 months, 55k OHLSV. What kind of history would I need if I switched to 1h, or at least 15m? And so I have 3 months, well, 6 months. for everything - about everything.
this is not enough history, you need a year at least 1m
that's not enough history, you need at least one year
To do what? What will you see there that you couldn't see in a month?
To do what? What will you see that you couldn't see in a month?
Well, it's about the difference between a month old baby and a year old baby.
And with a resistor, it's true)
In radio engineering it's even simpler, you don't need to know this stuff. Two-pole, four-pole, etc. - their parameters, and what's inside doesn't matter at all.
- A radio station on an armored train...
- Is it a tube radio or a transistor radio?
- I repeat, radio station on an armored train.