Discussion of article "CatBoost machine learning algorithm from Yandex with no Python or R knowledge required" - page 2

konorti
18
konorti  
Aleksey Vyazmikin:

Please upload the log after the CB_svod.csv script is running.

Please find attached,

I also attache the CB_Bat set file

Files:
Aleksey Vyazmikin
17733
Aleksey Vyazmikin  
konorti:
Please find attached

There is no error in the log. Send the file in the zip archive "output_test". I assume that there are no rows with a probability greater than 0.5. Very small sample.

Cols(2413) Rows(289) - test.

From these results, we need to select values with a high custom parameter — preferably 35% and higher, with a number of trades 15000 and more (the more the better). Optionally, other econometric variables can be analyzed.
konorti
18
konorti  
Aleksey Vyazmikin:

There is no error in the log. Send the file in the zip archive "output_test". I assume that there are no rows with a probability greater than 0.5. Very small sample.


I uploaded the whole Rezultat folder:

http://www.mediafire.com/file/m2uqcnkqu6ee8nb/Rezultat.zip/file

About sample size: Optimization was done on EURUSD 2016.01.01 - 2020.01.01. The highest trade number optimization which had more or less flat balance curve had around 1400 trades: 


Rezultat
Rezultat
  • www.mediafire.com
Compressed archives combine multiple files into a single file to make them easier to transport or save on diskspace. Archiving software may also provide options for encryption, file spanning, checksums, self-extraction, and...
Aleksey Vyazmikin
17733
Aleksey Vyazmikin  
konorti:


I uploaded the whole Rezultat folder:

http://www.mediafire.com/file/m2uqcnkqu6ee8nb/Rezultat.zip/file

About sample size: Optimization was done on EURUSD 2016.01.01 - 2020.01.01. The highest trade number optimization which had more or less flat balance curve had around 1400 trades: 



"Rezultat\RS_32\result_4_Test_CB_Setup_0_000000000" Probability<0.5, but one >0.5. What is the balance for the test sample?

Line of balance is good, but in this case it is important a greater number of transactions. The more examples, the more knowledge about the market.

konorti
18
konorti  
Aleksey Vyazmikin:


"Rezultat\RS_32\result_4_Test_CB_Setup_0_000000000" Probability<0.5, but one >0.5. What is the balance for the test sample?

Line of balance is good, but in this case it is important a greater number of transactions. The more examples, the more knowledge about the market.

I am not sure you are asking for the result for RA_32 from the CB_Svod.csv file (I also noticed that this was the only row which had all positiveresults): But still the mqh model was not generated for this. Do you have an idea why? Thanks for all the help.

Name_Dir Balans_Train Balans_Test Balans_Exam
model_Rezultat_RS_32_0 24552 985 34
Aleksey Vyazmikin
17733
Aleksey Vyazmikin  
konorti:

I am not sure you are asking for the result for RA_32 from the CB_Svod.csv file (I also noticed that this was the only row which had all positiveresults): But still the mqh model was not generated for this. Do you have an idea why? Thanks for all the help.

Name_Dir Balans_Train Balans_Test Balans_Exam
model_Rezultat_RS_32_0 24552 985 34

Now it is obvious that the script is working correctly. Models are created only those that have passed the filter, which by default is 3000 points. Lower this threshold in the script settings.

input double Filter=3000;//Minimum balance for copying and saving the chart
konorti
18
konorti  
Aleksey Vyazmikin:

Now it is obvious that the script is working correctly. Models are created only those that have passed the filter, which by default is 3000 points. Lower this threshold in the script settings.

Thanks! I will try further from the optimization part.
Aleksey Vyazmikin
17733
Aleksey Vyazmikin  
konorti:
Thanks! I will try further from the optimization part.

I am glad that You decided to reproduce the entire process described in the article. Good luck!

konorti
18
konorti  
Aleksey Vyazmikin:

Please upload the log after the CB_Calc_Svod script is running.

I cant do better :( from 2014-2020 on EURUSD

Aleksey Vyazmikin
17733
Aleksey Vyazmikin  
konorti:

I cant do better :( from 2014-2020 on EURUSD

Maybe try the settings from the article?

  • "Period": 8;
  • "Timeframe": 2 Minutes;
  • "MA methods": Linear weighted;
  • "Calculation price": High price.