Machine learning in trading: theory, models, practice and algo-trading - page 2055
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
Have you ever tried time series prediction?
it makes no sense
it doesn't make sense.
why?
Why?
Because the logic for the bot is to buy/sell, all the same forecasts are translated into signals
I should tell you right away that if you use the square of the error as an error function, nothing will work. The model will learn to predict the current barit makes no sense
The idea is that you can predict one bar forward, if you know the closing price of the bar on the daily timeframe, why not open a deal at the opening price and close them after passing, say, 50% of the distance from opening to closing? then you need to be at least 50% more accurate than the model
in theory you can predict for 1 bar forward, if you know the closing price of the bar on the daily timeframe, why not open transactions at the opening price and close them after passing, say, 50% of the distance from opening to closing? then you need to have at least 50% accuracy of the model
Regression models are more difficult to learn and are not needed
regression models are harder to learn and it's not necessary
I'm going on vacation to try it
sampling first, then splitting
you can do the opposite, the result will be the same because the sampling is random
I gave the function earlier, no one reacted
Max, do it right, as it should be and you'll get your 59% acurasi on returnees
first divide it into traine and test.
the train semplicate not semplicate, in short. do what you want !!! ))
the test do not sample because the test is a simulation of new data for the model, and the new data comes from the terminal without sampling !
When you do it this way, you'll get your 59% and it won't be 79% at all, but honestly.
Max, do it right, as it should be, and you'll get your 59% acurasi on returnees
first divide it into a track and a test.
the train, don't split it, do what you want !!! ))
the test do not sample because the test is a simulation of new data for the model, and the new data comes from the terminal without sampling !
When you do this, and you have to do it this way, then you'll get your 59% and it won't be 79% at all.
That's a great idea, how are you going to watch the error on the test without labels?
Relax, have a cofee, think.
How are you going to look at the error on the test without the marks?
What's with the marks? Why not?
I haven't watched your videos, did you come up with something non-standard there?
Check out the little ones ))))
Two of these for each data scientist!What's with the tags? Why not?
I haven't watched your videos, did you come up with something non-standard there?
Babies are cool. )))).
Two of these for each datasyst!!!Oh, man...