neural network and inputs - page 36
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
pdf doesn't want to attach, rar doesn't want to attach either. What do you need?
Explain:
four models were evaluated first and the result was
1. mlpe with AUC=0.924 and Acc=85.7%.
2. DT with AUC=0.877 and Acc=84.4%.
3. mlp with AUC=0.874 and Acc=81.7%.
4. svm with AUC=0.857 and Acc=82.4%.
i.e. ensemble of multilayer neural networks activated by different (random) initial values of weights showed better results than andomForest and decision tree?
No. The ensemble is better than DT,mlp and svm. The RF and ada values are given next and they are better.
would the difference in Acc between 85.7% and 89.4% give a significant improvement in the forecast?
I have a linear regression and a non-linear regression giving Multiple R values for e.g. gold of 0.95485489 and 0.97386429 respectively. I have not found any significant improvement in the predictive properties of the model in practice - in trading
The Ass=91% for the ada model. And this is very good. I didn't do regression. Or rather, I did, but I didn't like it.
What does the Multiple R value show? I haven't seen it before.
The Ass=91% for the ada model. And this is very good. I didn't do regression. Or rather, I did, but I didn't like it.
What does the Multiple R value show? I've never seen it before.
Multiple R is the multiple correlation coefficient.
Question - there are two methods. Using one gives prediction accuracy, for example, 1-3% better than the other - it will not give a tangible trading advantage of one method over the other.
Now if you divide the slope angle by the deflection, you get one value that fully characterises the trade. This can now be used as a fitness function for tuning.
You are the one who, sorry, "invented" the Sharpe indicator. A really good indicator, by the way.
Trythis teacher. (https://www.mql5.com/ru/code/903). You can't do better than that.
The inputs are whatever you like, you can even have OHLC.
It's not the Teacher, it's the underachiever. Like the joke.
A conversation on a trolleybus.
-Can you tell me when the stop is coming up?
-You follow me. -As soon as I get off, your last one.
In your case, you need a forecast at least three bars ahead. And that's regression.
And if you think you're a pioneer here, forget it. This direction is thoroughly trampled. Read more.
Good luck
It's not the Teacher, it's the doubletalker. As in the joke.
A conversation on a trolleybus.
-Can you tell me when the stop is coming up?
-You follow me. -As soon as I get off, your last one.
In your case, you need a forecast at least three bars ahead. And that's regression.
And if you think you're a pioneer here, forget it. This direction is thoroughly trampled. Read more.
Good luck
(Laughs)
There's no regression there, the regression you feed to the input in your example.
I've looked at "your" BBCI, it's no better and it's also glitchy.
Suggest the input data (excluding OHLC) and the teacher.
I'm not claiming anything, you asked, I offered.
You don't seem to have it figured out and regression is on your mind.
Keep stomping. Good luck.
Data can be fed into the input after spectral transformation of the input vector.
The task of the neural network in this case could be to predict the "future" of the spectrum. I have done a bit of research on this topic. I think there is a sense in such transformation, though it is a resource-intensive calculation. Here HERE I have described in more detail, some variants of application.