Machine learning in trading: theory, models, practice and algo-trading - page 1003
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
Hi, Misha!
Yes, it's time to rethink the efforts of all neuralnetworkers and their feeble hopes only on the tool itself. Nothing will help - neither forests, nor steppes - if the input data are not prepared.
And yes - there is no competition, there is a problem and there is general dumbing down.
If you know how to prepare the data, lay it out. Humanity will thank you.
For my part, I propose to look at the charts of EURUSD for 1.5 days:
On the bottom charts:
on the left - the number of real ticks in the sliding time window = 4 hours
On the right - the sum of increments and dispersion of this process.
As we see, in fact the dispersion directly depends on the amount of tick quotes in the sliding window.
There are reasons to suppose that if the sliding window size increases (for example, to 1 day), the process variance in the right side will be nearly const.
Conditionally such process can be considered stationary and neural network tools can be applied to it.
Alexander, your problem is that you look at kotir as an unsteady BP and nothing more. I already told you about it. Try to take the test for cfr 1.0 certificate for traders working for banks (I may have got the name wrong). You'd be surprised at how diverse the market is in general. One can't do without statistics of historical data here....
Misha, I'm not arguing - I could be wrong about something. But, I see a real crisis of the genre in this branch.
I have a cool thing in my hands - NeuralNet package for VisSim - and I am afraid to even touch it, because I see, that quite smart people are not able to do anything here.
If in my branch, in diffusion processes, I know a thing or two, here I have to still read and learn. What is there to learn? That "all gone, neural networks don't work..."? You need at least 1 person with a positive signal, even if it's +1% per month. It would really inspire a lot of people, me included.
.
If in my branch, in diffusion processes, I know, then here I still have to read and learn. And what is there to learn? That "it's all gone, neural networks don't work..."? You need at least 1 person with a positive signal, even if it's +1% per month. That would really inspire a lot of people, me included.
Please
The formula for calculating the error is shown in the header of the table. Let me explain on the last nnet example: 204/(204+458) = 30.8%, i.e. the model gave a total of 662 units, of which 204 were false.
The results are almost the same on 12 currency pairs, i.e. model performance is almost independent from the model and currency pair.
This result is obtained due to careful work with predictors whose predictive power changes very little when running a window of 500 candles over a file of 5000 candlesticks. The changes are within 5%.
PS.
I cannot show you the tester yet - it is stuck in the application of the tester for files over 1000 bars.
Purchased a non-growing network from some people. I am building a network similar to sigmoid. I have been testing on m15 for 3 years. I have not placed it for the last half year in testing in order to see how it will work. I set it up and everything is fine in the tester. I start to use real quotes and I am losing right away. I am in the red. I have tried it with different brokers. I read the thread, as I understand it, the neural network does not work in the markets?
How did you buy a cat in a poke? I sympathize.
How did you buy a cat in a poke? I'm sorry.
I used to buy another product from them, no problems. But here it started.
How did you buy a cat in a poke? I'm sorry.
It's only logical: trick-or-treaters are bound to be punished. ALWAYS.