Machine learning in trading: theory, models, practice and algo-trading - page 2790
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
Explained many times.
couldn't find it, they didn't write about R2, and this is important for me :)
Well, she wrote that she's in the eating and drinking category. So no solutions are on the horizon. Maybe at least one bright idea will slip through, it's already an accomplishment. There was some babbling about LSTM, then Agents. At the level of unconscious delirium so far. A set of words into a system is not yet formed.
We hope it'll rub off. He seems like a nice guy.)
Hopefully it'll rub off. He seems like a nice guy.)
She's getting used to it.)
Chel niche, it is not in the sense of human qualities. In the MO, the man understands, sometimes even something understandable not only to her rants))))
I couldn't find it, they didn't write about R2, which is important to me :)
Didn't write about R2. I don't use it. I don't remember the exact reason, it seems to be a measure for stationary rows, and we don't have such rows.
Yes, I've already banned him - it doesn't help.
Yes, sometimes he makes off-topic posts, recalling Maxim's posts here in the thread.
And often - on-topic.
For now, just ignore ...
Thanks for the response.
I've shown a similar approach before.
The essence of the approach is to search for analogues in history and compare them with the current situation/pattern....
There is a current pattern, we look for a similar pattern in history, look at how historical patterns ended and wait for something similar in the future...
This time I have completely given up analysing the price in its pure form because of the big noise, now I look for similarities purely in the curves of the indicator, in this case the bolinger bands.
The price is not taken into account, only three bolinger bands, the first 50 points are the current pattern, the one after the vertical line is the future.
Since the pattern is big, 3 curves of 50 points = 150 points, I reduced the dimensionality with PCA and took the first 25 components.
and proximity is looked for in the reduced space.
Here are some of the closest patterns found.
You can get more.
Well, it's nothing new, chaos and chaos...
This suggests that predicting the next value or the ZZ is not the best solution...
There is also an interesting hint, the figure shows that the price is kind of afraid to go below -0.5 for the first 50 points of the new data.
This is an interesting hint how you can and probably should make targets, and their appearance is quite different from what we are used to imagine....
p.s. visualise your data, it's useful, to say the least.
I have shown a similar approach before ...
....p.s. visualise your data, it is useful, to say the least.
I did a similar thing with prices about 4 years ago (channel through high/low). The pictures were similar. From all future variants (which are considered as a forecast) I built an arithmetic mean line. The forecast almost always turned out to be a horizontal line. I did not do it on the MO, but looked for minimal deviations from the template and variants in history. After such disappointment I started to use MO. But it is still there (horizontal or 50/50).