Machine learning in trading: theory, models, practice and algo-trading - page 2410
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
What to do when switching from range to range
This is just a thought in the form of pictures... I don't have any ready answers.
You can calculate the range relative to the last price, constantly shifting
Also there is a cool idea how to reduce prices in a more compact and repeatable form AMO this form of data should like
Pros :
1) Data in a much simpler form than raw data, which POSSIBLY will ensure repeatability and adequate learning
2) I think there is WHOEVER practically no information lost in such a conversion.
the algorithm is
1) I cluster prices using the dbscan (it is the most intellectual one) and is able to remove noise
2) save the average price of each cloud, as well as the number of points in the cloud
get the points of the clusters centers as prices at the top, and at the bottom how many points are in the cluster
or like this
I am pleased with myself ))) while the code is not finished)))
to compare the same pattern before and after transformation
I'm for some innovative approaches )
Another thing is that to dance from the stove does not mean to dance as if forever nailed to this very stove) And most discussions about SB on this forum look exactly like that, why they are already a little bit tiresome)
Well, there's no getting away from the SB - it's the very furnace from which our dances begin) Another thing is that to dance from the furnace - does not mean to dance as if forever nailed to this very furnace) And most of the arguments about the SB on the forum look like this, which makes them a little annoying)
I thought about non-standard preprocessing long and painfully, at that time nothing worked. Maybe I will think about it again, or I will find some material. Some good ones (but still crutchy). Use the residuals of a linear regression between the eurusd and the dollar index as a fic, regression residuals for the last n years for an instrument in the hope that the trend will continue.
index arbitrage
I thought about non-standard preprocessing long and painfully, at that time nothing worked. Maybe I will think about it again, or I will find some material. Some good ones (but still crutchy). Use the residuals of a linear regression between the eurusd and the dollar index as a fic, the regression residuals for the last n years for the instrument in the hope that the trend will continue.
Wait, but the standard dollar index has a known formula, right? So its logarithm is equal to the linear combination of the logarithms of six rates. That is, the residuals of the linear regression (for logarithms) will be a linear combination of the logarithms of the five remaining rates, no?
Wait, but the standard dollar index has a known formula, right? So, its logarithm is a linear combination of logarithms of six rates. That is, the residuals of the linear regression (for logarithms) will be a linear combination of the logarithms of the remaining five rates, no?
No, I mean the regression of the index on the eurusd. If you put all the tools instead of the index, not the fact that the same formula will work, I haven't checked it.
Not that I mind much (especially if it works). I just can't really see how or why it would work. Probably, it turns out that in the residuals of your regression all useful information is collected together and useless information is removed (about the behavior of the remaining currencies participating in the index).
Not that I mind much (especially if it works). It's just that I can't really see how or why it might work. Probably, it turns out that in your regression residues all useful information is gathered together and useless information is removed (about the behavior of remaining currencies participating in the index).