Machine learning in trading: theory, models, practice and algo-trading - page 64

New comment
 
Mihail Marchukajtes:
If the criterion for the point to appear is the future movement, then it is pure redrawing. That is, there is no point, the movement started, and the point appeared in the past. What should we classify? We need a fait accompli, not an event that depends on the future.....
And training for what? The classifier will predict these points right in the current moment.
 
Alexey Burnakov:

Again. In m hours the price will break 100 pips up or down. And so we model a trade with a take profit of 100 pips.

Or in m hours the price will be at least 100 points above/below. Then it simulates closing of a position in m hours. I do it this way.

Within m hours from what point? From the current hour, then the signal is the closing of each hour. I.e. one hour closed and gave a signal... It's not a buy or sell signal, it's just a signal and the Predictor classifies an hour if the price is going to be 100 pips higher or lower in 5 hours. Did I understand you correctly?
 
Once again, do not confuse the forecast with the classification. The forecast says that in 5 hours the rate will be 100 points higher or 121 points or 122, and the qualifier says that the current situation suggests an increase in the rate, and you don't know where to go, the main thing is to grow ... Let's not confuse the concepts...
 
Alexey Burnakov:
What is the training for? The classifier will predict these points right at the current moment.

Let me make it shorter:

The idea of Mihail Marchukajtes is not to use the classifier for predicting the future price direction, but to take some signal of the TA and classify its readings on whether it "lies" or "does not lie". I.e. the classifier can be used as a lie detector (polygraph) for some signal system based on TA.

 
Anyway, you can send me the xv file, I will train it and send you the binary model, and you check it on your own. How well it works... But it would be more correct to classify the signals of any system. For example the crossing of the wands.... So it will be more correct when working with the classifier. I think so.... Event occurred. In contrast to your event, the waving gets a face in the form of buy and sell, and your event does not have this face. Just an event... and nothing more....
 
Yury Reshetov:

I will try to formulate it shorter:

The idea of Mihail Marchukajtes is not to use the classifier to predict the future price direction, but to take some TA signal and classify its readings for "lying" or "not lying". I.e. the classifier can be used as a lie detector (polygraph) for some signal system based on TA.

Quite right Yury, quite right....... But if we take the waving, it will be a classifier of further price direction, because the waving follows the trend. The problem is that it goes back when it's not necessary, on pullbacks. The classifier will know that it's a pullback and will not let us open down and the price will stand further up. In your case it's a waving indicator to predict future price movements.....
 
Mihail Marchukajtes:
Once again, do not confuse the forecast with the classification. The forecast says that in 5 hours the rate will be 100 points higher or 121 points or 122, and the qualifier says that the current situation suggests that the rate will go up. Let's not confuse the concepts...

It would be desirable to use well-established terminology and not to impose your own homegrown one.

  • There are regression models that predict SIGNIFICANCE. For example, the value would be 1.2345.
  • There are classification models that predict, in particular, the direction, e.g. BUY/SELL. And in general, they predict a nominal (categorical) value.
  • Moreover, the model may be the same. But the target variable is different: in the first case it is a real number, and in the second case it is a nominal value.

 
Mihail Marchukajtes:
Within m hours from what point in time? From the current hour, then it turns out that the signal for you is the closing of each hour. I.e., an hour closed gives a signal... ...the signal is not impersonal buy or sell, it is just a signal and you use a predictor to classify an hour to see if the rate in 5 hours will be 100 pips higher or lower. Did I understand you correctly?

The signal is not impersonal.

Yes, from the current time. I have it to open a minute bar (model by opening prices). What I do is that my classifier gives out either 1 or -1 (or however I want to code it). 1 is a strong move up is predicted. -1 predicts a strong downward move. These are just a few bars, about 10% of all or even less.

 
Mihail Marchukajtes:
I may send me the cv file anyway, I will train it and send you the binary model and you may check it on your own. How well does it work? But it would be more correct to classify the signals of any system. For example the crossing of the wands.... So it will be more correct when working with the classifier. I think so.... Event occurred. In contrast to your event, the waving gets a face in the form of buy and sell, and your event does not have this face. Just an event... and nothing more....

I reread it several times, but I can't understand anything.

Especially confusing is the example of the dash. The point is that the same machka crossing - fast one crosses slow one from bottom to top, in the future it may lead both to the growth of price by the sought 100 pips and to its fall

 
SanSanych Fomenko:

It is highly desirable to use well-established terminology rather than imposing your own homegrown terminology.

  • There are regression models that predict SIGNIFICANCE. For example, the value would be 1.2345.
  • There are classification models that predict, in particular, the direction, e.g. BUY/SELL. And in general, they predict a nominal (categorical) value.
  • Moreover, the model may be the same. But the target variable is different: in the first case it is a real number, and in the second - a nominal value.

The classifier cannot predict here you are wrong. The classifier determines the current state of the system, and by determining it we conclude about the growth or decline of the rate.....
1...575859606162636465666768697071...3478
New comment
Reason: