Machine learning in trading: theory, models, practice and algo-trading - page 1926
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
and search and create, and search from what I've created )) and create from what I've searched ) and by magua and all sorts of things )
and then selects the best ones through selection and rearrangement, etc., etc.well cool, cool, but it's still a primitive functional search, and i'm talking about a full random search....
For example, this is the kind of thing your oscillator won't create -
if the candlestick at 6 o'clock was white and the price was lower than the price of the 6th hour candlestick before 12 o'clock, then it sat off the price of the 12th hour
You know what I mean? I'm not talking about multiplying the functions with each other, I'm talking about creating a complete random sampling of everything, without any indiscretions
You need to consider EVERYTHING! time, prices, levels, patterns, their sequences, everything, everything, and search, search, search
Well cool, cool, but it's still a primitive functional search, and I'm talking about a full random search....
This is the kind of thing your generator won't create.
if the candle at 6 o'clock was white and before 12 o'clock the price was lower than the price of the candle at 6 o'clock, then it sat off the price of 12 o'clock
You know what I mean? I'm not talking about multiplying the functions with each other, I'm talking about creating a complete random sampling of everything.
You have to consider everything! Time, prices, levels, patterns, their sequences all, everything, everything, and search, search, search
I don't think it works that way.
I don't think that's how it works.
What do you see as the problem?
What do you see as the problem? combinatorial explosion?
Sort of, there are infinitely more rules
like that, the rules are infinitely many
two words - dimensional reduction
That's what it's for.
You think I just got hooked on it.)
I already see from primitive tests that there is potential
two words - dimensional reduction
You think I just got hooked on this.)
I can already see from the primitive tests that there is potential
i realized what the flaw in my approach ... the criteria for entering the trades were determined by the mean and std of the cluster being trained, and the new data has a shift
we need to revise the criteria and that's it
try it... i'm kinda lazy for now )
I realized what was wrong with my approach. the criteria for entering trades were determined by the average and std of the cluster being trained, and on the new data the bias
we have to revise the criteria and that's it
it won't change anything, but keep trying.)
try it. I'm kinda lazy at the moment )
don't know how, yet...
==================================
i've done some training on yumap classes, visualized classes, you can see that over time, classes replace each other, especially where there is a strong concentration ("regularity") those systems not only die over time but even work backwards
Any idea how to do this?
What exactly?
what exactly?
Generally, the principle of generating fiches.
random...
A ficha can be represented as a log rule...
The size of the rule - randomly
content of a rule - random
generated 1000 rules - sent to the Defense Department as 1000 features
selected 1-5 good features if there are any, if not - we drop all of them.
selected features get thrown into "good features database".
and again we generate 1000 features, and so on
when there are over 1000 good features in the "base of good features", you will be able to use them to train a new model and see what you will get