NCA-AIIO Question 65
Single answerYou are comparing the performance of two machine learning models that predict housing prices: Model A and Model B. Model A has a Mean Squared Error (MSE) of 15.2, while Model B has an MSE of 12.8. Additionally, Model A has an R-squared (proportion of explained variance) value of 0.85, and Model B has an R-squared value of 0.78. Based on these metrics, which model would you choose for better prediction performance, and why?
- A
Model A, because it has a higher R-squared value, which indicates it explains more variance in the target variable.
- B
Model B, because it has a lower Mean Squared Error, which suggests it makes more accurate predictions.
- C
Model A, because its Mean Squared Error is higher, which indicates better generalization performance.
- D
Model B, because its R-squared value is lower, which makes it less prone to overfitting.
Show answer and explanation
Correct answer: B
Explanation
When comparing models using statistical performance metrics, Mean Squared Error (MSE) is a direct measure of prediction accuracy, with lower values indicating better performance. Although R-squared provides insight into how well the model explains the variance in the target variable, it is not as directly tied to prediction accuracy as MSE. In this case, Model B's lower MSE makes it the better model, even though its R-squared value is slightly lower.
- A. Incorrect.
While Model A has a higher R-squared value, the lower MSE of Model B indicates it produces more accurate predictions, which is typically more important in regression tasks.
- B. Correct.
Correct. Model B's lower MSE shows that its predictions are closer to the actual target values, making it the better choice for prediction performance.
- C. Incorrect.
This is incorrect. A higher MSE indicates less accurate predictions, not better generalization.
- D. Incorrect.
This is incorrect. A lower R-squared does not imply less overfitting. In fact, R-squared is more about the proportion of variance explained, not overfitting.