Article: Price forecasting with neural networks - page 17
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
I use a neural network expert. My result is in the attached file.
I think the result speaks for itself whether it is worth using neural network technologies in trading. I am doing an in-depth study of neural networks and I can say this is not the limit for these intelligent machines of the future :)
Where is the result?
I use a neural network expert. My result is in the attached file.
I think the result speaks for itself whether it is worth using neural network technologies in trading. I am doing an in-depth study of neural networks and I can say this is not the limit for these intelligent machines of the future :)
2 nsk: There is no attached file...
A very interesting thesis. Quote from the conclusion:
В заключение следует отметить, что качество прогнозов, полученных при помощи
нейронных сетей, выше, чем в любом из рассмотренных линейных случаев. Это позволяет
сделать вывод о наличии нелинейной зависимости между котировками валютного рынка.
Именно поэтому линейные методы регрессионного анализа не дали ожидаемых
результатов. Безусловно, теория линейных параметрических и непараметрических методов
прогнозирования изучена более подробно, чем относительно молодая (в ее современной
форме) теория нейронных сетей. Однако по итогам полученных результатов предпочтение
по праву отдается нелинейным методам.
A very interesting thesis. Quote from the conclusion:
To conclude that there is a non-linear relationship between currency market quotations."
Nothing to say, "deep thought." Actually, this so called dissertation is just a kind of term/diploma work
of a university graduate. And we all know how coursework/diploma theses are written.
"This allows
to conclude that there is a non-linear relationship between foreign exchange market quotes."
Nothing to say, "deep thought." In fact, this so-called dissertation is just a kind of term paper/diploma work
of a university graduate. And we all know how to write a term/diploma thesis.
Uh-huh. Exactly one test per method.
"This allows
to conclude that there is a non-linear relationship between exchange market quotations."
Nothing to say, "deep thought." In fact, this so-called dissertation is just a kind of term/diploma thesis
of a university graduate. And we all know how to write a term paper / dissertation.
I agree.
1) Volatility refers to the standard deviation of returns.
The classes within the trait are:
1. Low volatility
2. Medium volatility
3. High volatility
2) Average change per day is calculated as the sum of all changes for all days,
divided by the number of days.
Classes within the trait:
1. change per day is less than 50 pips on average
2. change per day more than 50 pips, but less than 150 pips
13
3. change per day averaging more than 150 points
Boo-ha-ha
Carl Linnaeus of the financial academy...
;)
I drew my conclusion after reading it in a non-diagonal way.
Here's an example...
The same SSA when moving the baseline for right-to-left decomposition by 1-2 bars
can drastically change the quality of the forecast from acceptable to who-knows-what.
So if you want to sell a program or write an article, you should take a good part of the chart and feel happy :-)!
We want to criticize the method - we show completely delusional forecasts.
... and the same LRF when forecasting a seemingly purely sinusoidal wave can draw a brutal nonsense.
And about the fact that he writes that there is no methodology for automatically grouping the components-
that's a bit strange, because that's what's implemented in statgroup's Caterpillar- and I've repeated it in my program.
.
... The rest of the "bugs" may be the same.
.
P.S.: I liked this coder:
http://www.nsc.ru/interval/Library/ApplDiss/Rodionova.pdf
I read the table of contents and laughed.
I should probably go to them to learn :-D.
Sorry, I don't play the market and don't know anything about forex. I'm a programmer and I'm writing a diploma on neural networks. The theme of my diploma is "Forecasting currency rates (euro and dollar) using neural networks", so I visited this site. So, you say that the prediction of currency rates at 70-75% - is from the realm of fantasy, while I want to say that if you learn NS, set parameters, filter the inputs to the NS, adjust the synaptic weights, etc., you can get the prediction accuracy of 90-97%. You can do it in Excel by installing Neural Package application. I did it :) There is even a manual for working with this application on the Internet, "Neural networks in MS Excel" by Fedotov V. H. The example for prognosis is explained there. It may help you too. Good luck to you.You should take a look at Reshetov's work. You may get up to 100%! Really, on the back-test... :)
Z.I. IMHO: you have to go for the price and pinch your crumbs, and in her actual predictions.... I don't believe it.