Machine learning in trading: theory, models, practice and algo-trading - page 880
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
NS usually processes the data from the first lines - i.e., the first lines have the old data in them, and the fresh data at the end, so that the last steps of training are done on them.
Thank you.
All predictors work on the bar opening - I don't even know how to identify those that peek - in theory their importance should be high?
Well, if everything is at Open price - then they should not peep, if other variants - then they peep.
Well if everything is at Open price - then it shouldn't peek, if other options do peek.
Can you test my dataset on some neural network, because I don't understand them yet? Built in that program does not want to learn more than 56% - maybe I'm doing something wrong, or the network is not suitable....
Can you test my dataset on some neural network, because I don't understand them yet? The built-in one in that program does not want to learn more than 56% - maybe I'm doing something wrong, or the network is not suitable....
Judging by the picture you have an error of about 8% (Accuracy 92%), not 56%.
You have a regression in your file, and you are testing a classification, judging by the description. Apparently you need a file with classification.
Judging by the picture, you have an error of about 8% (Accuracy 92%), not 56%.
Yes that file is wrong, I have already added predictors and made a classification of the target by a simple rule - if it is more or equal to 50 points, then 1 (for buying) and -1 (for selling), otherwise 0, and the columns for buying and selling are independent.
About 56% - so it's a neuron, and screenshots from the tree.Yes, that file is wrong, I have already added predictors and made a classification of the target by a simple rule - if it is more or equal to 50 points, then 1 (for buying) and -1 (for selling), otherwise 0, and the columns for buying and selling are independent.
About 56% - so it's a neuron, and the screenshots are from the tree.Well, if the tree works better, then use it. The NS is more difficult to set up.
So far only logic works (or maybe I'm testing wrong?), and I don't know how to use it.
If anyone wants to help check the workability of predictors, there are two files in the attachment - for buying and selling 3 column - the target, 1 and 2 I do not use, and the rest are predictors.
alglib has kfold, has anyone figured out how to work with it? almost zero documentation :)
Ah, I see, these methods teach automatically through cross-validation
Mm-hmm. The tools were given, but there's no manual. It would be nice to have help for each function, how to use, and with examples.
So far only logic works (or maybe I'm testing wrong?), and I don't know how to use it.
If anyone wants to help test the performance of predictors, there are two files in the attachment - to buy and to sell 3 column - the target, 1 and 2 I do not use, and the rest are predictors.
I tried file 1 and broke it into 3 parts:
Teaching
Predicted
Actual 0 1
0 28107 1244
1 3045 4119
Test 1
Predicted
Actual 0 1
0 5950 356
1 742 776
Test 2
Predicted
Actual 0 1
0 5945 333
1 779 769
Calculated on nnet with 10 neurons in the hidden layer (NS of Rattle package from R)
Worse than your scaffold, but not bad either. The second file will probably have the same results.