Machine learning in trading: theory, models, practice and algo-trading - page 2184
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
seriously...
Everything with me is 100 times easier.
I look at the channels as images with the algorithm, it's easier for me so far + it can be scaled to different TFs
Well yes, it is convenient to use them for aggregation. I think that it would be useful to build by ZZ - to describe the nature of price movement.
So, what are the cuts?
It is surprising, but I am not analyzing predictors efficiency globally - while I am doing it, and I see, that there is a dependence on strategy.
In general, now I continue to work on the selection of predictors, their usefulness for a particular TS. By analysis of quanta I select 100 predictors out of 4 and the result is comparable, but in the last case I already understand the validity of selection, because I select without training - just by analysis, as I wrote before, but the aim to improve the result - it is not always better than just search of grids with the same number of quanta for all predictors of models.
For example, you can select a predictor quite simply - take the predictors and look at how the response (percent 1) changes on each quantum of each sample (even if it's a training and test sample).
If the average delta modulo deviation of each quantum is small, then either the quantization is correct and the predictor is useful, and if not, then even a good predictor is useless, and probably it is not useful at all.
Fibonacci levels rule :)
Here it is different, I look at all standard TF 132 bars. (or 144 as suggested). After 1000 bars it seems to me that the memory of the series is lost. I.e. it is necessary to look at the state of large TFs. But so far I have not considered how to prepare data of different TFs.
I have another idea, the channel is built for a day and it is too boring for the large market to rebuild it :)
Is this the topic of the MO series?
Sure it is.
I'm thinking, instead of increments, SZ and balances, etc... why not make AMO draw channels.
I know what MO is. But what is AMO?
Scale with meaning, the small TF is a deciphering of the larger one.
I don't get the point....
it is a law of nature, it is an axiom...
I know what an ME is. But what is AMO?
Same thing ) machine learning algorithm AMO
It's just not always correct to write MO...
for example - I trained "machine learning" to class labels
I trained "machine learning" to class labels
That's not very fancy, is it?
And the AMO is fine.
didn't get the point....
it is a law of nature, an axiom...
The logic of extrema in manual logic))) Extremums on a low TF show the behavior inside a candle of a high TF. And if it is not uniform, i.e. chaotic, not trend-equal, for example, there is no signal for the beginning of the trend. The trend on the higher TF consists of trends upwards and downwards of the lower ones. This is a good confirmation.
The same is in the channels. But I can't tell you how to do this.
Is this a topic from the MO series?
It is already clear that it is a monkey's work. Actually, no one doubted it.
If your martyr doesn't lose this week - consider yourself lucky.
I'm sick of you. As a matter of fact, as always.
it is already clear that it is a monkey's work. In fact, no one doubted it.
If your marmoset doesn't get out this week, you're lucky.
I'm sick of you already. As always, in fact.
Come on, let's see that grail you've been calling a grail.
You write articles, lay out other people's articles here, fill people's brains, and you don't have a fucking clue what to do in the market...
And even more so, do not bullshit me, it's empty, I do not like to talk to you.
;)
Does anyone know the technology to represent several values by one with the possibility of converting back to several
with little or no loss of accuracy