Machine learning in trading: theory, models, practice and algo-trading - page 67

New comment
 
Mihail Marchukajtes:
Well, this is purely my observation, if the inputs are either 0 or 1, then 100 records will be unique. Anyway, you can send me a file, I'll train it by my method and send you the model, and you can see how it will work in the future... So... just curious!!!!
Training data file. The target is the far right, 3 class. You can remove all the zeros. I will then check for 1 and -1 trades.
 
I programmed exactly the same idea a week ago, described one pattern that +- works, and it is quite complex and the network traded only at the point where this pattern occurred, but I did not get any increase in the forecast at all, although I was sure that the trade will improve, I wonder why it happened so.... I had the same idea, why didn't it work for me?
 
SanSanych Fomenko:


There is no doubt that he does not know how to check the model, but I like the idea, and the rest is generally a matter of technique.

The idea is fresh, but it is not clear. Working or so.
 
Alexey Burnakov:
Training data file. Target - far right, 3-class. All zeros can be removed. I will then check for 1 and -1 trades.
Please make a semicolon instead of a comma as a separator between entries in the file, because I can't find it...
 
mytarmailS:
I programmed exactly the same idea a week ago, described one pattern that +- works, and it is quite complex and the network traded only at the point where this pattern was coming, but I did not get any increase in the forecast at all, although I was sure that the trade will improve, I wonder why it happened so.... I had the same idea, why it didn't work for me?
You just need to prepare the input data correctly....
 
Mihail Marchukajtes:
You just have to prepare the input data correctly....
How?
 
Mihail Marchukajtes:
Please make a semicolon instead of a comma in the file, because I can't open it properly...

So:

sep = ';', 

dec = ',', 

 
Did you give it away? And how long does it take you to train??? Well just wondering, there input data is about 30 or more... it will optimize all night long, and yes, in this case it is really long playing thing will turn out, quite possible that for years :-)
 
And in general, what is -1 in the output. It says in black and white either 0 or 1 minus 1 predictor is perceived as zero. So what to do?
 
mytarmailS:
How?
Send me the file with the data, I will apply it to you and you will see how it will work.... But your approach is absolutely correct, let's assume there is a pattern called BOLOT. Candlestick pattern, but it doesn't always work, so it too can be classified, those that work mark 1, those that don't mark 0 and when the hammer appears the model will tell you if it works or not. The approach is exactly right.....
1...606162636465666768697071727374...3477
New comment
Reason: