Machine learning in trading: theory, models, practice and algo-trading - page 781
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
heavily overfitted, pours on the OOS
I will train on a long interval, then I will show OOS
I also need to change the environment description.
So I understand that the periods of stochastics themselves don't change, but there is a weight inside, by which the signs are multiplied and then the activation function is executed?
So I understand that the periods of stochastics themselves do not change, but the weight inside is selected, by which the signs are then multiplied and then the activation function?
The target for the stochastics (3 RSI) are selected, i.e. there is no set of labels, yes
but it is trained not through the optimizer but through a full-fledged NS
Dear, advise, a way to classify if the numbers of bars for the inputs are known, but the reasons are not known.
What is the way to identify the patterns. To divide into two classes, where to enter and where not to enter?
Two vectors: one for longs, the other for shorts
Where to enter/be in position = 1, in others = 0.
The very problem is in the predictors. There should be ones that are relevant to the target.
If we don't have experience, we take rattle, 6 models, and the main thing is the full cycle: preparation of predictors, the model itself and evaluation of these models. If you prepare a file in excel, you can see all these results just at once, without understanding anything in R.
But there is a lot of material in this thread
Good luck.
PS.
Let's consider our regiment has arrived.
are enumerated targets for stochastics (3 RSI), i.e. there is no preset set of labels
Not through the optimizer, but through the full-fledged NS
What is a target, I should know?
I'm dealing with ARIMA. I realized that there are three steps:
1.Identification of a test model.
2. Estimation of parameters and checking for adequacy.
3. Prediction.
On the first point: it turns out that I need to make sure that the series is stationary, if not make a series of moments.
What is the target, I will know?
The target (label) is what is fed to the NS output during training (i.e. the value that it should then output)
and what is sent to the input is a trait (a feature, a predictor)
What are the targets, will I know?
I've been working with ARIMA for a while now. I understand that there are three steps:
1.Identification of a test model.
2. Estimation of parameters and checking for adequacy.
3. Prediction.
On the first point: it turns out that you need to make sure that the series is stationary, if there is no make of the series, a series of moments.
There is a function auto.arima that selects parameters automatically, and there are 3 (6) parameters, not one.
They check the residual from the model. There are special tests for this.
Two vectors: one for longs, one for shorts
Where to enter/be in position = 1 , in others = 0
The very problem is in the predictors. There should be ones that are relevant to the target.
If we don't have experience, we take rattle, 6 models, and the main thing is the full cycle: preparation of predictors, the model itself and evaluation of these models. If you prepare a file in excel, you can see all these results just at once, without understanding anything in R.
But there is a lot of material in this thread
Good luck.
PS.
Let's consider our regiment has arrived.
Thank you! Also, if it is not difficult where to read about 6 models, preparation of predictors, the model and its evaluation. I tried to work a little in R, but in years it is difficult to understand what is there.
there is a function auto.arima, which automatically picks up parameters, and there are 3 (6), not one.
They check the residual from the model. There are special tests for this purpose.
On the first point, I understood that you need to make sure that the series is stationary, if not decompose it into momentums. I checked it with ACF, CHAF and Dickey-Fuller test.
I even did ACF in MQL.
On the first point I realized that you need to make sure that the series is stationary, if not decompose it into momentums. To check the ACF, the CFC and the Dickey-Fuller test.
I even did ACF in MQL.
I don't know how to do it. I've already done it in MQL. Spit on mql - you won't get a huge number of different instruments and you'll be stuck with all sorts of stuff like acf and lots of other stuff. mql will be ready later when the model is working, but it may be impossible to get it at all. You'll just know in R that it's not working, but in µl you won't, because you lack the tool.