Discussion of article "CatBoost machine learning algorithm from Yandex with no Python or R knowledge required" - page 3

New comment
 
Aleksey Vyazmikin:

Maybe try the settings from the article?

  • "Period": 8;
  • "Timeframe": 2 Minutes;
  • "MA methods": Linear weighted;
  • "Calculation price": High price.
It seems that I had to set the capital from 10k USD to 200k USD so now I have at least better results: Score of 24 for 15k trades with 0.89 PF
 
konorti:
It seems that I had to set the capital from 10k USD to 200k USD so now I have at least better results: Score of 24 for 15k trades with 0.89 PF

In my code set the lot size equal to one. You consider the code as a template for experimenting with CatBoost.

 
Aleksey Vyazmikin:

In my code set the lot size equal to one. You consider the code as a template for experimenting with CatBoost.

thanks, it looks better now but from 40 seeds there was still no higher than 0.5. I try to do more seeds. Is it different if it goes 1 to 100 by 1 or 1 to 10000 with a step of 100?
 
konorti:
thanks, it looks better now but from 40 seeds there was still no higher than 0.5. I try to do more seeds. Is it different if it goes 1 to 100 by 1 or 1 to 10000 with a step of 100?

Do the quantization, and only then apply the seed. Each seed is different.

 
Aleksey Vyazmikin:

Do the quantization, and only then apply the seed. Each seed is different.

Thanks, I am not sure I fully understand you. Quantization and seeding is prepared in one step when setting it in the  CB_bat script, right?

Anyway during night some 200-300 seeds were generated with better results, also mqh files were generated. When I backtested during the training period the equity curve nicely picking up but when I test it during the test and exam period trades are rarely taken. Ma period was 96 so now I started again from the beginning. Switched to DJI30 (for a change) used period 8 and M2 and optimized the price and MA type only. This way much more than 15k trades are generated (I even reduced the length of  the period as the XXXCB_Save_pred.csv file is around 1.3Gb and 1 training cycle is  13 minutes. I set the seed parameter fro 1 to 10000  with a step of 100 which gives around 100 model. I hope there will be some result after this.

 
konorti:

Thanks, I am not sure I fully understand you. Quantization and seeding is prepared in one step when setting it in the  CB_bat script, right?

Anyway during night some 200-300 seeds were generated with better results, also mqh files were generated. When I backtested during the training period the equity curve nicely picking up but when I test it during the test and exam period trades are rarely taken. Ma period was 96 so now I started again from the beginning. Switched to DJI30 (for a change) used period 8 and M2 and optimized the price and MA type only. This way much more than 15k trades are generated (I even reduced the length of  the period as the XXXCB_Save_pred.csv file is around 1.3Gb and 1 training cycle is  13 minutes. I set the seed parameter fro 1 to 10000  with a step of 100 which gives around 100 model. I hope there will be some result after this.

I recommend that you first find the best way to quantize by going through the different options, and then go through the seed. Ideally, you should search for your own quantization settings for each predictor, and then combine the results. Perhaps I will write about this in the next article.

 

Thanks very much for sharing this great article and sample codes, while I have problem Running "_01_Train_All.bat" , 

scrrenshot as below:

Files in the folder:


could you please help to figure out where is the problem, or any step misconfigured?

very appreciated for your kindly support!

 
Di Wan:

Thanks very much for sharing this great article and sample codes, while I have problem Running "_01_Train_All.bat" , 

scrrenshot as below:

Files in the folder:


could you please help to figure out where is the problem, or any step misconfigured?

very appreciated for your kindly support!

You need to open the files exam.csv/test.csv/train.csv in a file reader (for example, EditPad Lite). Through the search, find and replace "inf" with 0, as an alternative to infinity. Unfortunately, in some indicators, non-numeric results can be obtained, this is rare.
 

If you want to use GPU you can comment out the frequency sampling line and add this to the array :

--task-type GPU

https://catboost.ai/en/docs/features/training-on-gpu


comment out this : // Train_All[25]=" --sampling-frequency "

I'm getting the following errors after I try to uncomment the model.  


Training on GPU
Training on GPU
  • catboost.ai
CatBoost supports training on GPUs.
 
dustovshio #:

If you want to use GPU you can comment out the frequency sampling line and add this to the array :

--task-type GPU

https://catboost.ai/en/docs/features/training-on-gpu


comment out this : // Train_All[25]=" --sampling-frequency "

I'm getting the following errors after I try to uncomment the model.


Please read the article carefully:

"

Select the model you like from the Models_mqh subdirectory of the directory in which our models were trained, and add it to the Expert Advisor directory. Comment the line with empty buffers at the beginning of the EA code using "//". Now, we only need to connect the model file to the EA:

//If the CatBoost model is in an mqh file, comment the below line
//uint TreeDepth[];uint TreeSplits[];uint BorderCounts[];float Borders[];double LeafValues[];double Scale[];double Bias[];
#include "model_RS_208_0.mqh";                 // Model file
"
1234
New comment