Machine learning in trading: theory, models, practice and algo-trading - page 1450

New comment
 
Of course I don't guess bars, but I try to keep the training result at least 80 on validation. By the way, I've done a lot of tweaking in Reshetov's work. I didn't change the minix search algorithm, but its organization has changed. For example I added cotrolny area and examine only those models that are giving good results just on the control, start search with maximal number of predictors, gradually reducing them, and also he had a chip of first run with random inputs, so look, I make these runs a few and choose the best and it is the starting point for basic optimization. In general, I finished it in the organizational plan so that is decent.... Java rules
 
Aleksey Vyazmikin:

So he refers there to already prepared data, the way of preparation of which he described here, and there:

"

Now let's form two sets of data, data1 and data2. In the first set as predictors will be digital filters and their first differences, and as the target - the sign of change of the first difference of the ZigZag. In the second set the predictors will be first differences of High/Low/Close quotes and first differences of CO/HO/LO/HL quotes, and the target is the first difference of the ZigZag. The script is below and can be found in thePrepare.R file.

"

It doesn't sound like we're talking about bars...

And, I think forex is not very convenient for MO, better go to Moex exchange.

I'm still doing other things, maybe this summer I'll have time to experiment with the MO.
 
elibrarius:
The bar color in Vladimir Perervenko's articles is very well defined with an accuracy of about 0.8. And it is easy to repeat. I did it again, including with naked quotes, the results were 2-3% worse than with digital filters.
It is strange that your results are so bad. Although much depends on the instrument and the TF.

I want to correct. I never predicted the color of the bar. Only classification with target ZZ and different variants of predictors. And the best results are obtained using large ensembles with different aggregation methods.

Another important experience is that the more complex/refined the preprocessing, the simpler the model solves the problem.

Good luck

 
Aleksey Vyazmikin:

So he refers there to already prepared data, the way of preparation of which he described here, and there:

"

Now let's form two sets of data, data1 and data2. In the first set as predictors will be digital filters and their first differences, and as the target - the sign of change of the first difference of the ZigZag. In the second set the predictors will be first differences of High/Low/Close quotes and first differences of CO/HO/LO/HL quotes, and the target is the first difference of the ZigZag. The script is below and can be found in thePrepare.R file.

"

It looks like we're not talking about bars...

And, I think forex is not very MO-friendly, better go to the Moex exchange.

Aleksey Vyazmikin:

So he is referring there to already prepared data, the way of preparing which he described here, and there:

"

Now let's form two data sets - data1 and data2. The first set will contain digital filters and their first differences as predictors, and the sign of change of the first ZigZag difference as a target. In the second set the predictors will be first differences of High/Low/Close quotes and first differences of CO/HO/LO/HL quotes, and the target is the first difference of the ZigZag. The script is below and can be found in thePrepare.R file.

"

It doesn't sound like we're talking about bars...

And, I think forex is not very convenient for MO, better go to Moex exchange.

Alexey, I completely agree with the last thought and you can mark this day with a pencil, because the archives are restored, the account is activated, I start working on Si. :Q=)

 
Vladimir Perervenko:

I want to correct. I never predicted the color of the bar. Only classification with target ZZ and different variants of predictors. And the best results are obtained using large ensembles with different aggregation methods.

Another important experience is that the more complex/refined the preprocessing, the simpler the model solves the problem.

Good luck

I hadn't really thought of the last thought in that way, but it's interesting. Now everything fits together, it remains to add that the data itself should have at least the slightest relation to the target. Well it is stupid to predict the weather according to the quotations of the euro. Agree!!!! And those who do this and get good results do not understand that this is just a coincidence, a very good coincidence in a million others, it just fell out. Checks are easy. You can't repeat the result.)

 
This is a reference to the recent epic of pseudo-programs....
 
Grail:

52% is not nonsense, but the truth uterus and by the way if you watch it is not such a bad result, quite tradable, with annual SR ~1

80% is some kind of humor, or near-market gimmicks.

Hmm, good, so I'll consider this improvement as an achievement.

And what about the possibility to trade it - yes, it is possible to put a pause just below the open price and close it after exceeding ATR 61,8% and the mathematical expectation may improve.

 
Mihail Marchukajtes:

Alexei, I completely agree with the last thought and you can mark this day with a pencil, because the archives are restored, the account is activated, I start working on Si. :Q=)

I do not know whether to rejoice for you or not - it seems that you were in stable remission and now you are with us again :)

I can't let go of this activity, alas...

 
elibrarius:
I'm doing other things for now, maybe in the summer I'll have time to experiment with MO again.

Didn't anything work out with ZZ for you?

 
Aleksey Vyazmikin:

I don't know whether to be happy for you or not - you seemed to be in stable remission, and now you're with us again :)

I can't let go of this activity, alas...

I've never stopped trading, not for a single day. I'm not allowed to idle, I just don't communicate here. I don't know why I joined the blog, but I've been trying to get my masters here. It's not right, so I decided to join the team back...
1...144314441445144614471448144914501451145214531454145514561457...3699
New comment