0

For example, in linear regression, we can say that the relationship between X and Y is linear. When X increases, Y also increases. On the other hand, for non-linear regression (Power model as an example), we can say that the relationship between X and Y is exponential relationship. When X increases, Y increases exponentially.

Is it possible to describe (draw) the relationship between input and output in Machine learning? Is there any Algorithm, approach, etc. to describe, shape, or draw the type of relationship between X and Y in Machine learning (Supervised)? ANN, SVM, Tree, etc ...

kjetil b halvorsen
  • 77,844
Yazan Alatoom
  • 15
  • 1
    Have you tried plotting the prediction of Y for various values of X? – dimitriy Dec 10 '22 at 01:59
  • @dimitriy You can't always find the relationship from the plotting. Sometimes when you have many X and no clear trend in the data then you can't say the relationship. In the first, if you know the shape of the relationship why you don't only perform empirical regression – Yazan Alatoom Dec 10 '22 at 12:42
  • 1
    You could also look into Shapley values decomposition. – dimitriy Dec 10 '22 at 19:33
  • 1
    The regression coefficients give you the change in expected Y from one additional unit of X, keeping all other variables constant. The graph of predicted Y varying X and keeping other variables constant does something like that, though you can plot the change in predicted Y to make it more apples to apples. You do know the relationship since you have a model. Note that I am talking about predicted Y, not Y. – dimitriy Dec 10 '22 at 19:36

0 Answers0