Machine learning in trading: theory, models, practice and algo-trading - page 3348
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
Itdoesn't matter what the profit margin is. Itmatters what the classification error is afterwards.
Because of this approach you "correctly" classify potentially losing trades. In reality, the situation is much worse not only because of the spread. In a real EA to get from "correct" classification to a profitable system remains a problem, as it is not surprising.
First the markup is made as profitable as possible. Then "reliable" examples are resampled and filtered based on model errors, the rest is marked as rubbish. Because it is clear that there will never be such an ideal trade as with the initial grail markup (without spread it will be almost a grail). Profitability falls to some level, stability on new data grows. A balance between this and that is chosen.
It seems logical and not so vague as others justify their TS.
I have described the easiest variant for understanding in the article, you can check it yourself, the core of the algorithm is simple.
First, the markup is made as profitable as possible. Then "reliable" examples are resampled and filtered on the basis of model errors, the rest is marked as rubbish. Because it is clear that there will never be such an ideal trade as with the initial grail layout (without spread it will be almost a grail). Profitability falls to some level, stability on new data grows. A balance between both is chosen.
It seems logical and not so vague as others justify their TS.
The easiest variant to understand is described in the article, you can check it yourself, the core of the algorithm is simple.
I took a quick look at the article.
I have singled out a certain basic premise from the very beginning, on which everything else is based:
If we train the model many times on random subsamples, and then test the quality of the prediction on each and sum all the errors, we get a relatively reliable picture of the cases where it is actually wrong a lot and the cases it guesses often.
Totally disagree.
Any cross validation cannot, by definition, improve the quality of the model. Cross validation allows you to compute a more VALID error value at the expense of a set of statistics. All. and the resulting classification error may or may not have anything to do with the prediction on the external file.e. in real trading.
The quality of prediction by a model is determined by the set of predictors for a particular set of labels and has nothing to do with the model. Before modelling, the question must be answered: do the predictors and their labels fit together? It is impossible to answer this question with the help of a model, and you are trying to do it.
It is impossible to answer this question with a model, and you are trying to do so.
What do you want to answer it with?
and what do you want to answer with?
Old topic and written many times over.
Old topic and written many times over.
support by viewing, varnishing, subscription of the developer of the package R rusquant
It's the same thing
In your article there is no graph for the "forward" mode of the tester, by which one can really judge the model.
By the way, you use mashki, no matter what the difference with the price, and you should be careful with them, because under certain conditions of testing models by your own testers, as it is not funny and contradicts the whole TA, mashki look ahead. Using the "forward" mode, if there is looking ahead, you will get a big discrepancy in the results between the forward and the main plot.
rusquant
The site says thatinteraction with API Tinkoff, Finam and Alor is supported. Has anyone looked into it?
The site says thatinteraction with Tinkoff, Finam and Alor APIs is supported. Has anyone looked into this?
No...
I am not interested in his business, I just used rusquant package for many years and I am grateful to him for that.