0

I have trained Gradient Boosting classifier, but when I validated the model on completely new data, the resuls were, due to totally different data, poor.

I have sample data from production process and my supervisor says it is normal that the errors in production process change rapidly (e.g. in time when there are new software upgrades). So she advised me to develop self-learning algorithm from the one I have already trained.

When I was googling the solutions, I found only general approach to the topic, but no real instruction to get me to the solution.

Could anybody help how to do?

I am afraid if this is available with my GB classifier, but I tried several algorithms for the data and this one was the best.

Thank you.

  • This is a coding Q&A site. You need to come up with your own code and submit specific questions about aspects of the code which are not working, or you can't finalize, or you are trying to optimize. You can't just ask a question and hope people will give you the whole code, for that you have Google (If you google for GB you will find some code on github). Regarding `when I validated the model on completely new data, the resuls were, due to totally different data, poor`: this sounds like a typical case of overfitting, so you need to add regularization. – Max Mar 03 '21 at 07:33
  • thanks, Max, this helped me. I was not aware of overfitting. – ZuzanaTelefony Mar 03 '21 at 07:46

0 Answers0