Apartment-prices: Residual distribution
The distributions of residuals for both models are different.
The residuals of random forest
- They are centered around zero so the predictions are, on average, close to the actual values.
- The skewness indicates that there are some predictions where the model significantly underestimated the actual values.
The residuals of linear-regression:
- They are splitted into 2 separate normal-like parts, located about -200 and 400, which may suggest the omission of a binary explanatory variable.
Random forest residuals seem to be centered at a value closer to zero than the distribution for the linear-regression, but it shows a larger variation.
