Machine learning in trading: theory, models, practice and algo-trading - page 617
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
The answer to your question: classification. Signals buy/sell/pass.
As for the images, it is not clear what kind of lines.
Screenshots of the MetaTrader trading platform
AUDUSD, H1, 2018.01.28
RoboForex (CY) Ltd, MetaTrader 5, Demo
I agree, the pictures are clippings from the chart.
About the classification, is this in the same neural network can be applied?
Model in understanding of this audience is a set of dataset parameters (columns, variables) + set of mathematical methods (functions) + result (function response).
I understand that you are talking about dataset.
The length of dataset affects the quality and speed of learning (rows). The quality of parameters (columns) affects the quality of the prediction
I do not know what to do with the length of the dataset. In my case, one and the same model can be 500 bars long, and in a day it will be 200 bars long. Suppose I make a script that for every hour will give me datasets, but they will be different length, how can I push them into a neural network? I think that just the length of model is also a quality parameter for neural network...
https://charts.mql5.com/17/376/audusd-h1-roboforex-cy-ltd.png
I agree, the pictures are clipped from the graph.
As for the classification, it can be applied in the same neural network?
From the images it is not clear what is the data? Where is the forecast? What do the vertical lines mean? In general, as they say, there is no legend on the chart.
Basically yes, in the structure of neural network model you will need to change the output layer (number of neurons, activation function) and transform data for training into classes (for example indicator signals, normalized price increments).
From the images, it is not clear what kind of data? Where is the forecast? What do the vertical lines mean? In general, as they say, there is no legend on the chart.
Basically yes, in the structure of neural network model you will need to change output layer (number of neurons, activation function) and transform data for training into classes (e.g. indicator signals, normalized price increments).
According to the picture, the vertical lines show the model itself, everything after the right vertical forward. So the forward behaves differently, and with the help of a neural network I want to find a possible explanation for the forward behavior. I want to shove the data for the neural network into the price increments of the model.
I do not know what to do with the length of the dataset. In my case, one and the same model can be 500 bars long, and in a day it will be 200 bars long. Suppose I make a script that for every hour will give me datasets, but they will be of different length, how can I push them into a neural network? I think that just the length of model is also a quality parameter for neural network...
Your dataset is a table where columns, for example, are opening/closing prices and rows are time moments, bars.
In a elementary neural network architecture it receives one row after another as input, and for each row it should output a result which after comparing it with a benchmark the optimization function will "train" the neural network.
If you are feeding it several strings at once, it is a time series, which means that it should be feeding it data according to a certain algorithm.
I should add: study articles on the subject, I think many questions will become clearer.According to the picture, the vertical lines show the model itself, everything after the right vertical forward. So the forward behaves differently, and with the help of a neural network I want to find a possible explanation for the forward behavior. I want to use data for the neuronet for price increments of the model.
I started to write an article for paired trading... but I got stuck because of my laziness and lack of understanding how to do it correctly :)
Your dataset is a table where columns, for example, are opening/closing prices, and rows are time moments, bars.
In a elementary neural network architecture, at its input it receives one row after another, and for each row it must produce a result, comparing which with the benchmark the optimization function will "train" the neural network.
If you feed several strings at once it is already a temporary series, which means that you should feed it according to a certain algorithm.
Sorry for the dumbness, my knowledge of neural networks is probably too superficial. So, for example, I have a sequence of clones prices over 100 bars and a forward of clones prices let it be 5 bars. 100 inputs go to the neural network and 5 outputs go to it. But now the next sample an hour earlier has a sequence of e.g. 200 bars and the forward also has 5 bars. The third sample an hour earlier has 250 on the input and 5 on the output. How to make such a neural network? Everywhere there are examples describing that the input data is in the same amount.
I started to write an article for doubles trading... but got stuck on my laziness and misunderstanding of how to do it right :)
Hello, I do not have a paired method, although I came through my acquaintance with paired. Today I use portfolio trading, and only because I create a portfolio for my convenient stage (model) at any time. And when I noticed that portfolio series in principle do not differ from any market series I had a desire to find possible regularities for the same stages (visually similar) using neuronet.
Maxim, you're running your neural networks on monopair, aren't you? Have you ever thought about the fact that you can create a convenient row, which may then show itself better on the forward? After all, in fact, let's say a headshoulders figure, for example, does not occur very often, but imagine that you can do it every hour...
Sorry for the dumbness, my knowledge of neural networks is probably too superficial. So, for example, I have a sequence of clones prices over 100 bars and a forward of clones prices let it be 5 bars. 100 inputs go to the neural network and 5 outputs go to it. But now the next sample an hour earlier has a sequence of e.g. 200 bars and the forward also has 5 bars. The third sample an hour earlier has 250 on the input and 5 on the output. How to make such a neural network? Everywhere there are examples describing that the input data is the same.