Linear models can handle non-linear relationships thanks to data pre-processing. But how close can they get to more sophisticated models?
—
Some weeks ago, I published a post on LinkedIn.
The post was based on the following figure, comparing the predictions made by two models: Linear Regression, and CatBoost.
The gist of the post was that a gradient-boosting model like CatBoost seems to provide a more “reasonable” interpretation of the relationship between a predictor and the target variable (namely, house condition and house price).
Indeed, many relationships in nature are non-linear.
The post received some objections, among which the following comment stood out for the large number of likes:
This started a discussion, and I found the following comment (written by the same author…