ForumSections

Machine learning in trading: theory, models, practice and algo-trading - page 93

New comment
Mihail Marchukajtes:
Fuck, it pisses this forum off that you can't attach a simple rar archive.... It's so lame....
Watch what you say. You can attach:(.gif .png .jpg .jpeg .zip .txt .log .mqh .ex5 .mq5 .mq4 .ex4 .mt5 .set .tpl)
And when you spin it, enough txt note on csv I will send you the formula by which you can achieve a 100% generalization .....
Karputov Vladimir:
Watch what you say. You can attach(.gif .png .jpg .jpeg .zip .txt .log .mqh .ex5 .mq5 .mq4 .ex4 .mt5 .set .tpl)
Why can't I use rapeseed? If the archives, then let all the archives... For example, I do not have zip... what to do ????
Mihail Marchukajtes:
Well, why can't I have it??? If archives then let all archives... For example, I have no zip... what to do????
zip is a Windows standard. Called in explorer by right-clicking on the file/folder.
Another thing is if we agree and don't use data lag..... Hmmm... It's interesting to try. The bottom line is this. We take the predicts, multiply, divide, subtract and add them to each other, get an output variable, make it output one or zero. We shove it into a predictor, each into its own program, and see if the program can build a model.... Hold on, let me try it :-)
Karputov Vladimir:
zip is a Windows standard. Called in Explorer by right-clicking on a file/folder.
I call it by right-clicking on a file/folder... OK, not the point....
Well, back to our game. Took and stupidly multiplied, divided, added and subtracted line by line input data, got values, made of them the output in the form of 0 and 1, and as a result the model is built with 82-95% generalization saying that there is a relationship between input data and output. That is, it once again proves that the data for TC to take from the ceiling just silly. The input data must be somehow connected to the output...... In the case of forex we are trying to build a model of the ideal output, but such input data simply does not exist, so the output should not be ideal, but it is suitable for making a profit, like this....
So??? Was anyone able to figure out my data????
Vizard_:
Conditions as I understand no one reads (It is allowed to use any data manipulation), so I will not torture. In fact, everything is simple.
Just need to take lags from A6, apply a simple formula seventh less than fifth and get 100% on both samples. Thank you all. Good luck...
Too bad you said your answer early. You could have tortured the learning methods.
Alexey Burnakov:
It's a pity you answered early. You could have tortured the methods of training.
It's not about the training methods. Using lag in predicates will not allow to build a sufficient model. After all, a line is fed to the model, and if the given line is used, the model simply won't see it, so the task is obviously a failure.
1...8687888990919293949596979899100...3748
New comment