Machine learning in trading: theory, models, practice and algo-trading - page 2509
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
For example, I take standard deviation, accumulation/distribution and stochastic component of OI, Delta and Volume data series and use them to make a forecast...
In accumulation/distribution I'm confused (for some reason) by the possibility of strong influence of someone's input or output, which will strongly distort the local picture
I feel that I will incur the wrath of the local IO intellectuals/alumni, but I will risk asking a question. Who thinks what indicators (other than built-in or the most popular ones) are the most interesting for the forecast? For my part I'm currently experimenting with combination of exponential moving average (DEMA) with Kalman filter and fast Fourier transform (separately). But first I make predictions using a neural network. Once again I clarify, I'm not asking for application results (otherwise the nice girl will again spew a bucket of slop).
Write here about what you're doing, what are your thoughts, etc..
Anybody can help.
Professionals are not here, there are 5-10% of practitioners, and the rest are yawners and babblers...
There are no professionals here, there are 5-10% of practitioners, and the rest are yawners and babblers.
There are no professionals, there are 5-10% of practitioners, and the rest are yawners and blabbermouths.)
And professional practitioners are most likely almost all sitting quants in the funds, there is nothing for them to do here)
100%
In the accumulation/distribution I'm confused (for some reason) by the possibility of someone's strong influence of entry or exit, which will greatly distort the local picture
No this indicator is only for that would interpret the volume to a greater extent, to obtain from a simple volume of the curve indicator. Something like that!!!
And to know the accumulation and distribution of the delta would also be nice!
Only the euro, the Swiss, the yen and the dollar (if you believe the link) "somehow" float freely (among the more or less liquid ones)... many are inflation-linked (austral, canadian, new zealand, pound) - their own targets and their own policies (there is little mathematics there) - just think of Fischer for general development
p.s.
It is better to model microeconomics or economic theory, but not macro (although there is everything in the interest rates)... or better not to model and monitor cme summaries (although not completely informative) or others...
Starting small is logical. But even a simple minority game model (bar with less people you're at an advantage), in case of small complications of initial conditions gets immediately curses of dimensionality, lack of resources, and if you consider averaged parameters, curses of not model accuracy)
Cleaned up the posts, second time I write)
In the R.mqh file, the names of the vector and matrix variables started to give an error when compiling. Rename them to other names and everything will work. I used vectr and matr.
The editor highlights these words in blue as a data type like int, double. Apparently reserved words for new types.
In short, all in vain, with MO the market can not cheat.
I found the signs and the target, whose class distribution is shown in the first figure.
The accuracy on the test and training katbust models trained using this dataset was 93%
The second figure shows the balance and equity graph of the target trade:
The third figure shows the graph of balance and equity trading on the signals of the trained catbust model:
So, ladies and gentlemen, disperse.
Neuralists here are of course authoritative, and know how to squeeze the most out of the data in the fight against overtraining, but in my opinion, it is the input data that is the main problem. Any oscillator (standard or self-written), and any continuous curve does not contain the regularity of price, so we can't teach the mouse.
I see the following way: trends and their waves. The wave length, the interval of time of the wave, the speed of the wave, a comparison of these parameters with the previous ones, the size of the excess of the previous extremums (trend movement), the distance to the nearest non-crossed extremum in the past, ... ...there are a lot of things that can be compared. I think there are regularities here, that is what you can indulge in with your grid,
or you can think for yourself.
I had such thoughts too, that's why I took up Fourier, but after spectrum cleaning and reverse transform there are more questions than answers. It turns out that we remove small influences, and large ones stop noticing local changes. I tried to apply it to the NS, but of course it did not lead to anything, from the point of view of common sense. I have not figured out some other way to distinguish waves.) The piece-by-piece comparison is of course valuable, but it is not for the MO, here all the regularities very quickly end.