Databricks Machine Learning Associate Question 256
Single answerYou are training a regression model to predict house prices. The target variable (house price) is heavily skewed, so you decide to apply a natural logarithm transformation to the target variable before training the model. After training, you calculate the Root Mean Squared Error (RMSE) on the transformed scale. How should you interpret this RMSE value in the context of the original house prices?
- A
The RMSE value directly represents the error in predicting house prices on the original scale.
- B
The RMSE value must be exponentiated to represent the error in predicting house prices on the original scale.
- C
The RMSE value represents the error on the log-transformed scale and cannot be converted back to the original scale.
- D
The RMSE value should be squared to represent the error in predicting house prices on the original scale.
Show answer and explanation
Correct answer: B
Explanation
When the target variable is log-transformed, the RMSE is computed in the log scale. To interpret the error in the original scale of the target variable, the RMSE must be exponentiated. This step ensures that the error estimate is meaningful in the context of the original problem, such as predicting house prices.
- A. Incorrect.
This is incorrect because the RMSE is calculated on the log-transformed scale, not the original scale. It cannot directly represent the error in the original house prices.
- B. Correct.
This is correct because when the target variable is log-transformed, the RMSE is computed in the log-transformed space. To interpret it in the original scale, the RMSE must be exponentiated.
- C. Incorrect.
This is incorrect because while the RMSE is calculated in the log-transformed space, it can still be converted back to the original scale through exponentiation.
- D. Incorrect.
This is incorrect because squaring the RMSE would not provide the correct interpretation of error in the original scale.