Machine learning in trading: theory, models, practice and algo-trading - page 987
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
For the appetizer.
A run of the same Expert Advisor with the same settings as above, but with a longer time interval.
This is the whole value of all these nice pictures.
The picture should prove the idea whose meaning is ONLY about the future behavior of the Expert Advisor.
Sanych, don't jump to conclusions, you've screwed something up.
The number of transactions over a longer period of time is less
That's why I use my tester, not the MT tester - for some reason it has a lot of grails. In your tester, at least you know reliably - what and how it does. Yes, and information from the test can get much more and any, and easier.
In my opinion, it's not very good to write such things on MT forum, considering that the backtester in MT is one of the main problems, that binds people to mql and the entire terminal.
Sanych, don't jump to conclusions, you messed up something.
The number of deals for a longer period of time is less
I have run it twice more: the graph is similar, but the numbers are a bit different.
I ran it two more times: the graph is similar, but the numbers are slightly different.
Conclusion
Making a decision to buy or sell at random will not provide a stable profit
Found a problem with the classification:
For example, if 2 columns = 0, trying to do a softmax of them gives random classes:
m=matrix(0,ncol=2,nrow=100)
max.col(m)
[1] 1 2 2 2 2 1 2 1 2 2 2 2 1 1 1 2 1 1 2 2 1 2 1 1 2 2 2 2 1 1 2 2 1 2 2 1 2 1 1 1 1 2 1 2 1 1 1 2 1 2 1 1 2
[54] 2 2 1 2 2 1 1 2 2 2 2 1 1 1 1 1 2 2 2 2 2 1 2 2 2 2 2 2 2 1 2 2 2 1 2 1 2 2 2 1 1 1 2 1 2 1 1
It's in R.
I stumbled upon it by accident when it turned out that the prediction results all = 0.
It is better to do so (in case the 1st column means "expectation" and not trade command)
max.col(m,ties.method = "first") # by default ties.method = "random"
[1] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
>
Better yet, if the classes have equal value, then refuse to classify. And just in case, it's better to do it line by line.
Found a problem with the classification:
For example if 2 columns = 0, trying to do a softmax of them gives random classes:
m=matrix(0,ncol=2,nrow=100)
max.col(m)
[1] 1 2 2 2 2 1 2 1 2 2 2 2 1 1 1 2 1 1 2 2 1 2 1 1 2 2 2 2 1 1 2 2 1 2 2 1 2 1 1 1 1 2 1 2 1 1 1 2 1 2 1 1 2
[54] 2 2 1 2 2 1 1 2 2 2 2 1 1 1 1 1 2 2 2 2 2 1 2 2 2 2 2 2 2 1 2 2 2 1 2 1 2 2 2 1 1 1 2 1 2 1 1
It's in R.
I stumbled upon it by accident when it turned out that the prediction results all = 0.
It is better to do so (in case the 1st column means "expectation" and not trade command)
max.col(m,ties.method = "first") # by default ties.method = "random"
[1] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
>
Better yet, if the classes have equal value, then refuse to classify. And just in case it is better to do it line by line.
Afternoon
The problem may not appear if the data are properly prepared. When and in what calculations did you have this problem? I wonder. Or it is an artificially created condition?
Good luck
Good afternoon
The problem cannot appear if the data are prepared correctly. When and in what calculations did you have this problem? I wonder. Or is it an artificially created condition?
Good luck
As SanSanych said: garbage in input = garbage out. I also added zeros, which then got canceled.
Well, I made a correction to keep zeros as zeros, for similar cases in the future.
Predictors are different, so I came across that the NS did not learn anything and gave all zeros at the output. And predictions due to randomness in the conversion turned out to be non-zero.
As SanSanych said: garbage in input = garbage out. I also added zeros, which then got canceled.
Well, I made a correction to keep zeros as zeros, for similar cases in the future.
I see. Good luck
Questions from a newbie. Please advise on methods of applying machine learning. For example, a trader has found a certain pattern in the market. Suppose it is a GP (head-shoulders) pattern. Variants:
Questions from a newbie. Please advise on methods of applying machine learning. For example, a trader has found a certain pattern in the market. Suppose it is a GP (head-shoulders) pattern. Variants:
Machine learning is built on signs (patterns/features) that will distinguish the event. Accordingly, you need to specify what you should look at, and the MO algorithm will try to find some regularities in what is shown and to work out the rules of behavior. All the answers to the other questions follow from here. And accordingly, the more observations, the more accurate the rules will be on a longer period of history.