Machine learning in trading: theory, models, practice and algo-trading - page 492

 
Andrey Kisselyov:
Perhaps the platform is paid...

With respect.

Mm-hmm, Quik.) But if you really work, it pays off in one transaction. And trades in a month - at least 60. In general, it does not matter.

If you meditate in front of the monitor, or do not go in, of course, in a year there won't be anything left.)

 
forexman77:

Do not write the name, you can be banned. Everything is clear with the name. There are only two brokers, as far as I know.


I do not get banned here for Otkritie :) I don't like BKS because of their forex. Once I went to their office to make a presentation about their analytical complex, they gathered their managers there and the local director came... I did everything properly, but it was not useful to them :) but they sent their customers a link to me, which is also respectable

i have fun as i know how

 
Yuriy Asaulenko:

I even know who you mean). They don't, if you choose a normal broker. With programming will have to retrain)). !5 thousand to learn and start earning is over the top.

Perceptron MLP - 6 layers. I have already written, the structure is 15,20,15,10,5,1. This is only for longs. For shorts I have another similar one.


I'm studying C# and it will be useful in any case, not in trading, but in other spheres.

 
forexman77:

I'm studying C#, it will come in handy in any case, not in trading, but in other spheres.

C# is good. But you also need (better) C++ - the organization of interaction with terminals is not the last task. The DLL's have far from simple functionality of type - received-transferred.

 

It seems to me that we need a network that will look for patterns graphical and to them some "classics" to attach. But what to use for this, maybe Kohonen?

 
forexman77:

It seems to me that we need a network that will look for patterns graphical and to them some "classics" to attach. But what to use for this, maybe Kohonen?

The usual MLP is fine... Although you only have 3 neurons yet, and you already have maps.

Didn't you tell me what environment you work in? R?

I must say, I don't like R - it's a big poorly systematized dump of everything and anything. Maybe there are some diamonds there, but you'll have to search for them.)

 
Yuriy Asaulenko:

The usual MLP is fine... Although you only have 3 neurons yet, and you already have maps.

You didn't tell me what environment you're working in? IN R?


Yes, I do everything in mql. What should I do in R, if I don't know it)

First I copied one layer, then another, etc.

 
forexman77:

Yes, I did everything in mql. What do I do in R, if I do not know it)

First I did one layer, then another and so on.

I don't know much about it - it takes me about an evening to master R. More in-depth in three.)) But it's really not so easy to deal with packages and how to adapt and apply them to your own tasks.
 
Yuriy Asaulenko:
There's not much to know - R itself is mastered in an evening. (More in-depth in three)). But it's really not easy to deal with packages and how to adapt them and apply to your tasks.

It's a different language, you have to learn it in three evenings, probably it won't work...?

In algotrading there are a lot of examples and books in Python, R, much less in C#, even less in C++.

I read that it's easier to understand C#, then I found out that Python. I will use C# for the time being.)

 
forexman77:

Yep) It's a totally different language, you have to learn it in three evenings, probably you can't...?

In algotrading there are a lot of examples and books in Python, R, much less in C#, even less in C++.

I read that it's easier to understand C#, then I found out that Python. I am going to train in C# for now)

Actually, the language itself is only 30-40 reserved words). They all (languages) are about the same and have different nuances. Each language has a lot of libraries, and they are really unique in the way they are used. It's useless to study them too - you just have to take the ones you need.