Forecasting currency exchange rates using a neural network - page 4

New comment
 
nikelodeon:
To be honest, I don't trust forecasts either, but classification, that's more interesting...

Where do you want this classification to go afterwards? The decision to buy or sell is made on the basis of the relationship to a certain class, i.e. a forecast of the expected price movement is made.
 
Integer:
Forecasting is not necessarily extending the line into the future. Any means of analysis here is about predicting whether price will go up or down.


Of course, in the end, in any case the financial market operation is forecasting, but we know what we are talking about. An uncharacteristic forecast and a forecast based on a pattern that does not appear very often are fundamentally different things.

 
Integer:

Then where does this classification go? A decision is made whether to buy or sell on the basis of the relationship to a certain class, i.e. a forecast of the expected price movement is made.

Belonging to a class, ultimately a buying class or a selling class at a particular point in time, means specifically "selling" or "buying". Without all the nuances.

Predicting a rate means a certain line into the future, which must coincide with the future line of that rate given the nuances of the trade.

These are specifically different subjects of "forecasting".

 
C-4: But still, at the moment we enter the market, we expect the price to change in the direction of our entry in the future, and this is the forecast.
Working in the financial market, in any case, is a forecast, but a forecast based on a found pattern and a daily forecast are essentially different forecasts.
 
LeoV:

Belonging to a class, ultimately a buying class or a selling class at a particular point in time, means specifically "selling" or "buying". Without all the nuances.

Predicting a rate means a certain line into the future, which must coincide with the future line of that rate given the nuances of the trade.

These are specifically different subjects of "forecasting".

There you go, you couldn't have said it better +100500 classification essentially just predicts the direction at a given time, without the nuances of how long, how deep, etc. I especially like comrades who call future levels.... It's so funny to look at them :) Although they sometimes show it on the news :) Anyway, in this exact neural network technique, we predict only the direction at a point in time and not more......
 
KID:
Hello, could you tell me if it is possible to use the STATISTICA package to predict the exchange rate for 10-15 days. This is an assignment for a course project. I found a similar example: http://www.statistica.ru/statportal/tabID__32/MId__141/ModeID__0/PageID__354/DesktopDefault.aspx, predicting gold exchange rate, but I'm a bit confused how to feed the input data to the neuron, could you help, if not with advice, at least with understandable reference literature?

It might be useful https://www.mql5.com/ru/articles/1506
 
nikelodeon:
essentially only predicts the direction at a given point in time, without the nuances of how long, how deep, etc.
The prediction window is there anyway, no matter classification, regression... So both how long and how deep everything is taken into account (at least it is possible), just not in much detail because it is unnecessary information, and neural networks do not digest it well.
 
LeoV:
A financial market operation is a forecast in any case, but a forecast based on a pattern found and a forecast every day are different in nature.

Leonid, either the thought was expressed incorrectly, or in an attempt to make it clear you are confused yourself.... "Yebar forecast" like any other time series forecast is also made on the basis of patterns found, for there is nothing else to rely on. Although I suspect that was put in the meaning. Price prediction and direction prediction? Of course, they are different things.
 
Figar0:
Leonid, either the thought was expressed incorrectly, or in an attempt to explain in an accessible way you yourself are confused.... "Yebar prediction" like any other, as applied to a time series, is also made on the basis of a found pattern, for there is nothing else to rely on. Although I suspect that was the point. Price prediction and direction prediction? These are of course different things.
Yes, I agree, there are inaccuracies in expressions here. Of course, the calculation is done on every bar, but if in the first case we calculate the forecast on every bar, then in the second case if no pattern is found on a new bar, then there is no forecast.
 
nasdaq:

Might come in handy https://www.mql5.com/ru/articles/1506

Thank you. Quite a lively forum.

1234567
New comment
Reason: