Machine learning in trading: theory, models, practice and algo-trading - page 70
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
Hello!
I am trying to run a convolutional network from mxnet packagehttp://tjo-en.hatenablog.com/entry/2016/03/30/233848, but not really, or rather not clear how to run it with "our" data, namely in the form of strings, because the network works mostly with pictures and data takes the form of a multidimensional array with matrices, in short, if anyone understands and knows how it should be run, I would be very grateful for an example of the network, say, "iris"
In the heat of a new topic did not notice my post, +10 pages in one day is not a joke)
But still ... I ask for help, I really need it, I will publish the results
Listen! Nobody uses Reshetov classifier here except you, most of us use R programming environment, it's much more flexible approach in all directions than to use some separate product... If you explained properly what and how, then I think each of us would be able to implement both algorithm and trade back test, you know? Just explain properly what to do and how to do it. I told you that a week ago I implemented the same concept, but it did not work, you say it's all bullshit, you need to properly prepare the data, so how to prepare them I ask for the third time? and you will have an implementation and a backtest, and in different variants from each of us....
How do I know what you will program there on R to prepare the data?
Hello!
I am trying to run a convolutional network from the package mxnet http://tjo-en.hatenablog.com/entry/2016/03/30/233848 but not quite, or rather not clear how to run it with "our" data, namely in the form of strings, because the network works mostly with pictures and data takes the form of a multidimensional array with matrices, in short, if anyone understands and knows how it should be run, I would be very grateful for an example of the network, say, with "iris
Whatever you explain we'll program, it's not clear what we're talking about...
If you are an expert in MetaTrader I also know that there is no difference between the optimizer and the trading robot. How do I know what you'll program into R to prepare data? I work with Reshetov's optimizer and it suits me fine. It's just a tool, now you have to find decent inputs and deo in the hat, and considering that optimization speed has increased, I think it won't be hard to do it....
It is useless to persuade people to switch to another software. It is psychologically difficult, I know this from my own experience and I have observed it many times on others. For example, when I worked in one organization, they installed new computers and launched Windows. But people didn't quickly learn Word and Excel, but started MS-DOS and used Lexicon to fill in all the documents, including tables.
In order to start a mass migration to other software, it is necessary to demonstrate a specific result, for example in the form of a profitable signal. When I created the AfterEffects Expert Advisor, I also ran the signal for it on the demo. The users saw the profit and began to download the Expert Advisor. At present, AfterEffects' optimization pages on my website are the most visited according to statistics, although the signal has been disabled for a long time. Apparently, someone has run the Expert Advisor in trading, earned profit and gave advice to others.
The same needs to be done with jPrediction. Build a fully automated bundle of jPrediction with MetaTrader, get profits at least on the demo, run the signal, make an instruction for users. And then people will start coming over.
What is there to explain? I'm not a developer of neural networks, I'm a user of them......
OMG....
when i wrote that i did the same thing as you but my result was zero, you said you had to prepare the data correctly what do you mean can you explain?
The same is needed with jPrediction. Build a fully automated jPrediction bundle with MetaTrader, get profits at least on the demo, run the signal, and make instructions for users. And then people will pick up.
The modern standards of jPrediction are not a software at all, and to compare it with R is in oooh oooh... you just need a lot of imagination.
jPrediction is one of the classifiers, and it's very important to have the environment in which this wagon is located to be able to compare them.
Even more important is something else.
To have a big enough set of tools for the preliminary preparation of the initial data. In addition to this, it is important to have a large enough set of tools to be able to evaluate the result.
And now, sorry, you're just advertising another trinket... You're confusing people...
By the way, I have laid out the data. Maybe someone will try to do it? I will separately post the out-of-sample set for trade assessment. There instead of -1;0;1 will be values of differences between prices with an interval of 3 hours. And you can calculate the expectation of the trade on the predicted signals.
I will try, train the model within a week. Then you can even check on fronttest, on yet unknown continuation of that file. But there is a nuance that the algorithm I work with is built on two classes, on three will be a problem with my fitness function to evaluate the model.
Right now I have two possible classes in the model, with a classification result of 0 or 1 for sell/buy. I'll apply the same code for your three classes, but with the output scaled to (0;0.5;1), but that's not the best approach. It would be better to make 3 outputs in neuronka corresponding to three classes, and take the result of classification as the output with the largest value. I don't know which of these ways would work best on your data, I'll make both of these models, I wonder which one will give the best result.
I will try, train the model within a week. Then you can even check on fronttest, on yet unknown continuation of that file. But there is a nuance that the algorithm I work with is built on two classes, on three will be a problem with my fitness function to evaluate the model.
Right now I have two possible classes in the model, with a classification result of 0 or 1 for sell/buy. I'll apply the same code for your three classes, but with the output scaled to (0;0.5;1), but that's not the best approach. It would be better to make 3 outputs in neuronka corresponding to three classes, and take the result of classification as the output with the largest value. I have no idea which of these methods would work best on your data, I'll make both of these models, I wonder which one will give the best result.