Machine learning in trading: theory, models, practice and algo-trading - page 1623
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
Is there an account monitor?
I have not started trading yet because the Fullblood machine is still being tested for reliability. In fact, due to the fact that the futures ends today, so the last week or two quotes are working purely technically and any system starts working, even the so-so systems. Tomorrow we move to a new futures. At the beginning of the quotation everything always goes hard, there and we'll see...
I see, I am in the test too, but it is too long))). Now pictures with arrows no, there is only this (profit in pips).
Sure it may be just by coincidence)).
It's not like that...
One model (a neural network) does not give the right result. It can learn something, but it's not enough. That's why I build 20-25 models with different inputs. Now I have 25 models signaling at the same time, and each model's opinion is considered with certain weight in final forecast. Calculation of one model takes about 0.5-0.7 seconds totaling 15-20 seconds + we should prepare input date for 25 models, that is a lot of work for every minute) The answer can be reduced to 1-3 seconds if I use multithreading correctly in python, but I haven't done that yet.
I train models separately, in normal mode, i.e. dataset is collected from history period from one year and then training as usual.
What's the target?
It's not like that...
One model (a neural network) does not give the right result. It can learn something, but it's not enough. That's why I make 20-25 models with different chips on inputs. Now I have 25 models signaling at the same time, and each model's opinion is considered with certain weight in final forecast. Calculation of one model takes about 0.5-0.7 seconds totaling 15-20 seconds + we should prepare input date for 25 models, that is a lot of work for every minute) The answer can be reduced to 1-3 seconds if I use multithreading correctly in python, but I haven't done that yet.
I train models separately, in normal mode, i.e. dataset is collected with history period from one year and then training as usual.
It feels like you are training a model every minute. Getting the result of a trained model usually takes a thousandth of a second.
What's the target?
no target, all equal
What do you mean ? I mean what is Y ?
It's not like that...
One model (a neural network) is not enough. It can learn something, but it's not enough. That`s why I build 20-25 models with various inputs. Now I have 25 models signaling at the same time, and each model's opinion is considered with certain weight in final forecast. Calculation of one model takes about 0.5-0.7 seconds totaling 15-20 seconds + we should prepare input date for 25 models, that is a lot of work for every minute) The answer can be reduced to 1-3 seconds if I use multithreading correctly in python, but I haven't done that yet.
I train models separately, in normal mode, i.e. dataset is collected with history period from one year and then training as usual.
Hmm, "0.5-0.7 seconds of calculation" is too much for MLP, maybe you teach and then calculate, on small datasets with a sliding window?
Let's go better in order:
1 What is the raw data (ticker(s), timeframe)
2 What is the size of the training dataset (1k,10k,100k...)
3 What kind of features
4 What are the targets
5 What kind of grid
Enough to start ...
Hmm, "0.5-0.7 seconds of calculation" is a bit much for MLP, maybe you teach and then calculate, on small datasets with a sliding window?
Let's go in order:
1 What is the raw data (ticker(s), timeframe)
2 What is the size of the training dataset(1k,10k,100k...)
3 What kind of features
4 What are the targets
5 What kind of grid
is enough to start with...
And you still need to shine the lamp in your eyes))