Discussion of article "Gradient Boosting (CatBoost) in the development of trading systems. A naive approach" - page 10

 

Model as coded in python

Here you can see the code for the model. Take note of MA_Period, Look_Back, etc. Then Look at the python code tester profit curve. Then look at the MT5 inputs, settings and strategy tester results.

the profit curve(using predictions by the model)

The Settings The inputs The MT5 Strategy Tester Results

[Deleted]  
Busingye Tusasirwe #:

I seem unable to reproduce the results from the python tester. The MT5 tester is not reproducing the results for the same period in python tester.

Otherwise, I ported the model as explained. 

I put cat_model.mqh and cat_trader.mql5(compiled to .ex5).

But results look different.

Hello, there may be a difference between how the model was parsed when the article was written and how it happens now. CatBoost could have changed the code logic of the final model in the new versions, so you'll have to figure it out.

It seems to me that there is a high probability that this could be a problem.

 

I've made a few changes:

I changed the code to save the mqh according to the chart time of the data.

I changed the mqh to be different for each timeframe, so that it's possible to have all the timeframes trained and ready to use in the EA.

I changed the EA to use all the trained files for analysing and generating signals.

All the files are attached for your perusal, if possible.

If you could improve the code I would be grateful.

The strategy and also the model training need extreme improvement, if possible I appreciate help.

I've made a few changes:

I changed the code to save the mqh according to the graphical time of the data.

changed the mqh to be different for each chart time, so that all chart times can be trained and ready to use in the EA.

I changed the EA to use all the trained files for analysing and generating signals.

I have attached all the files for you to analyse if possible.

if you can improve the code i would be grateful.

the strategy and also the training of the model needs extreme improvement, if possible help thanks.