Machine learning in trading: theory, models, practice and algo-trading - page 2030
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
retrained the same way as the others
There's a lot of information on google.
Here's a good one.
https://towardsdatascience.com/gate-recurrent-units-explained-using-matrices-part-1-3c781469fc18
) I have a problem with English))), and there are almost no articles in Russian.
and the ones we have, it is not clear how to update the weights of the connections. to be precise, it is not clear how the weight that goes from the output of a neuron to its input is updated.He probably made the same mistake in his tester.
I don't think so. Remember, he started testing with other data, looking for some boolean function or something... ...and everything was fine.)
Found a bug in the tester, finally. It all made sense now.
closed the topic ) in a hurry, the Lamba is postponed.
What's the conclusion? it does not work or what do you need to rewrite?
CatBoost will swallow this data - it works well with large files. I just didn't understand what is the target...
If you will prepare data in the form of csv with the target, I can start at myself.
Any suggestions? haven't decrypted the result yet.
My options, from the input information to determine the result of the transaction +/- or predict the exact value.
I don't know how, I would like to get information on favorable parameters: entry time, hold/exit time, direction, SL, TP.
Even more interesting, make a constructor of strategies. I have all data in the table, we just need coefficients for volume, direction, and we may construct any system, reverse, with martin, etc.
And I managed to create a system where TP is 2-3 times greater than SL, but there is another problem - 20%-25% of winning trades, and I can not properly train the model, which would sift out unprofitable entries.
ihnho, SL=TP is very visual, honest and convenient to compare, but from observations SL>TP is better, although it can be attributed to martingale
And I managed to create a system where TP is 2-3 times greater than SL, but there is another problem - 20%-25% of winning trades, and I can not properly train the model, which would sift out unprofitable entries.
You can try to express the target in a more complex way in the form of four parameters at once
Suppose we decide to buy...
and the grid doesn't just tell us to buy or sell.
it tells us
at what price to buy, at what price to close, in how much time to buy and in how much time to close
you can add a stop loss
I don't think so. Remember, he started testing with other data, looking for some boolean function or something... and he was fine.)
So what's the conclusion? it doesn't work? or do you need to rewrite something?
I had loss-making trades leaking into the tester under the guise of profitable ones. I rewrote it - the chart was reversed, no way)
There's no way to make it work on the upside
There's a Boolean function with 5 features, 4 lines... I'm sorry... that chase it for hundreds of iterations. It's not even funny. It's like what you did, but there's very little there. I don't know, though.
) I have trouble in English))), and there are almost no articles in Russian.
And those that are available, it is not clear how to update the weights of connections. to be precise, it is not clear how to update the weight that goes from the output of a neuron to its input.https://www.mql5.com/ru/articles/8385
not the fact that there is a good implementation )
I'll pass on Russian
we can try to express the target in a more complex way in the form of 4 parameters at once
Suppose we decide to buy...
and the grid doesn't just tell us to buy or sell
it tells us
at what price to buy, at what price to close, in how much time to buy and in how much time to close
you can also add a stop loss
Why not take the basic strategy of any format and build the training file according to it? In this case, there won't be any problems with the target one. And in the end, you want to make four grids instead of one.
Especially interesting how you want to get an answer from the grid at what price to buy, given that the price should be minimally rationed? It can be done but only in case of the initial reference point, I think it is the next bar after the training and as a result the obtained target will need to be further transformed to get a specific price figure.
Any suggestions? I haven't deciphered the result yet.
My options are to determine the result of the trade +/- or predict the exact value from the information about the entry.
I don't know how, I would like to get information about favorable parameters: entry time, hold/exit time, direction, SL, TP.
Even more interesting, make a constructor of strategies. The table has all the data, you only need to substitute the coefficients for volume, direction, and you can build any system, reverse, with martin, etc.
I wish you success :)
Do you need regression? I do not have much experience in such models.
I am familiar with this concept - there are people who do it - the question is what method to create strategies - in the engine itself...
Imho, SL=TP is very clear, fair and convenient to compare, but from observation SL>TP is better, although it can be attributed to martingale
The market is volatile, fixed TP/SL are not always effective in similar chart conditions. That's why it's better to be tied to specific entry points, that way the learning should be better according to the idea.