Probabilistic neural networks, packages and algorithms for MT4 - page 17
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
Please tell me, esteemed "torturers" of neural networks, should I spend my time on neural networks to solve the following problem:
In MQ4 I am writing an EA with input variables like start time, duration, SL, TP, position number in the array, ..., then I fill the array with strings - "symbol name" in the body and define in the algorithm, exactly buy or sell, depending on, for example, the following condition ((Futsi/Dow at 17:00 (MSC) the previous day) - (Futsi/Dow at 10:00 (MSC)) I start optimizer with steps on all variables ( by time, let it be, step=5m,...) and I wait for the results..., and accordingly further analyze the results.
So, will using neural networks help me to find such patterns?
When I heard about networks, I thought - there's something in it, maybe it will be useful for me...,
and after spending about five hours researching it, I decided to get some advice...
What software and how to use it? Will it make my research easier?
Please tell me, esteemed "torturers" of neural networks, should I waste my time fiddling with neural networks, to solve the following problem:
>> Definitely not.
Thanks for the unambiguous chorus answer, but I must have been too ambiguous and I think I misunderstood the question. I was interested in the possibility of technical implementation of such ideas with . As for strategies like " Predicting gold price by analyzing slippers and weather will not work. :)" - it's called fundamental analysis, price dynamics of slippers is a component of inflation, meteorological threat to oil producers - price growth, etc. Another thing is that it is difficult to obtain and analyze such information before everyone else, so we can proceed from the postulate that "the price reflects everything" and the main task is "not to jump into the last carriage of a departing train". The idea of fully automatic trading seems to me utopia, moreover I analyze only one tool - I think it's like playing chess with only one piece (maybe not a very beautiful allegory, I hope it's clear what I mean...). So less flubbing, closer to the topic. Above I wrote that one of the conditions for opening a position is a correlation between instruments, in this case between the British securities and American, I have not mentioned that the pose opens accordingly for GBP / USD. This example is simple and MT tester is enough for me to check this assumption (half year run for 2-3 hours on one instrument from the array). But it gets worse, there is a more global idea on the same principle. As for the neural networks, it seemed to me that feeding the inputs only, for example, the previous low and guessing the present one is a fun for mathematicians, feeding the inputs with macroeconomic data + current instrument prices, at the moment of news release, is another matter. I even suspect that those neural networks rumoured to be used by Big "players" should work like that. They probably need a staff of economists, programmers, a bunch of neural hardware (boards), and an algorithm to interpret "analog" data (Bernanke speech, etc.) into understandable numbers. I have no illusions that one or a few enthusiasts can do it, so try to "jump if not into the first, then at least not into the last carriage". For example, the amount of money supply can be regarded as constant (the change is tracked by inflation data), the money is not kept in stockings, it flows from papers to the raw materials, metals, ..., so imho it makes sense to look for imbalances, and intersessionary. For this purpose one should investigate a lot of information. Try to find a "local" gap in comparison to the main trend. If you're interested, you can continue. Sorry, running away...
What did it say about how to actually do it?
And then there were the pros and cons of the neuron species.
Thanks for the unambiguous chorus answer, but I must have been too ambiguous and I think I misunderstood the question. I was interested in the possibility of technical implementation of such ideas with the help of neural networks. Regarding strategies like "Gold price prediction by analyzing slippers and weather is impossible. :)" - it is called fundamental analysis, price dynamics of slippers is a component of inflation, meteorological threat to oil producers - price growth, etc. Another thing is that it is difficult to obtain and analyze such information before everyone else, so we can proceed from the postulate that "the price reflects everything" and the main task is "not to jump into the last carriage of a departing train". The idea of fully automatic trading seems to me utopia, moreover I analyze only one tool - I think it's like playing chess with only one piece (maybe not a very beautiful allegory, I hope it's clear what I mean...). So less flubbing, closer to the topic. Above I wrote that one of the conditions for opening a position is a correlation between instruments, in this case between the British securities and American, I have not mentioned that the pose opens accordingly for GBP / USD. This example is simple and MT tester is enough for me to check this assumption (half year run for 2-3 hours on one instrument from the array). But it gets worse, there is a more global idea on the same principle. As for the neural networks, it seemed to me that feeding the inputs only, for example, the previous low and guessing the present one is a fun for mathematicians, feeding the inputs with macroeconomic data + current instrument prices, at the moment of news release, is another matter. I even suspect that those neural networks rumoured to be used by Big "players" should work like that. They probably need a staff of economists, programmers, a bunch of neural hardware (boards), and an algorithm to interpret "analog" data (Bernanke speech, etc.) into understandable numbers. I have no illusions that one or a few enthusiasts can do it, so try to "jump if not into the first, then at least not into the last carriage". For example, the amount of money supply can be regarded as constant (the change is tracked by inflation data), the money is not kept in stockings, it flows from papers to the raw materials, metals, ..., so imho it makes sense to look for imbalances, and intersessionary. For this purpose one should investigate a lot of information. Try to find a "local" gap in comparison to the main trend. If you're interested, you can continue. Sorry, running away...
o_O About the slippers - it's called basic neural networking.
Please tell me, esteemed "torturers" of neural networks, should I spend my time on neural networks to solve the following problem:
In MQ4 I am writing an EA with such input variables as start time, duration, SL, TP, position number in the array, ..., then I fill the array with strings - "symbol name" in the body and define in the algorithm, exactly buy or sell, depending on, for example, the following condition ((Futsi/Dow at 17:00 (MSC) the previous day) - (Futsi/Dow at 10:00 (MSC)) I start optimizer with steps on all variables ( by time, let it be, step=5m,...) and I wait for the results..., and accordingly, further analyze the results.
So, will using neural networks help me to find such patterns?
When I heard about networks, I thought - there is something in it, maybe it will be useful for me...,
andafter spending about five hours researching it, I decided to get some advice...
What software and how to use it? Will it make my research easier?
It seems to me that time spent on this probably isn't serious enough to draw conclusions.
There are people who've been working on NS for 10-15 years.
---
to get into it and understand, NS, what it is all about, a neural network!
I think if you're going to do it, you need to gather information first.
Read publications - books - articles - lots of them on the internet, GOOGL
---
there's a lot of software out there.
You might want to look at the most common packages.
look at statistica, neurosolution, they're pretty cool.
statistica is nice because there are releases in Russian
then there's brain
I think there are others.
For trading, the more closed package is NeuroShell.
I personally don't like black box - I prefer to write NS in C
---
for the skeptics
I remember :
I don't like cats - maybe you don't know how to cook them.
and the main objective is not to jump into the last carriage of a departing train.
The main task: to jump into any wagon of the train going in the right direction. Otherwise, you'll have to jump out at a stop-loss, or be kicked out of the carriage by a margin call - without money and in the middle of the tundra.