MLS-C01 Question 240
Single answerYou are building a machine learning solution for a retail company to predict product demand. You have trained three models: XGBoost, a neural network, and a linear regression model. The following metrics were recorded during evaluation for each model:
- XGBoost: Training time = 2 hours, MAE = 1.2, Engineering cost = Medium
- Neural Network: Training time = 10 hours, MAE = 1.0, Engineering cost = High
- Linear Regression: Training time = 10 minutes, MAE = 1.5, Engineering cost = Low
The company prioritizes faster training time and lower engineering costs over achieving the lowest error. Which model would best meet the company’s requirements?
- A
XGBoost
- B
Neural Network
- C
Linear Regression
- D
None of the models meet the company's requirements
Show answer and explanation
Correct answer: C
Explanation
The company's priorities are focused on fast training times and low engineering costs over achieving the lowest possible error. While linear regression has a higher MAE (1.5) compared to other models, it is the fastest to train (10 minutes) and has the lowest engineering cost, making it the best choice according to the company's stated requirements.
- A. Incorrect.
While XGBoost achieves a relatively good MAE of 1.2 and has a medium engineering cost, its training time of 2 hours is longer than necessary given that a faster, lower-cost model exists.
- B. Incorrect.
The neural network achieves the best MAE of 1.0 but is not suitable for the company's priorities due to its long training time of 10 hours and high engineering cost.
- C. Correct.
Linear regression has the fastest training time (10 minutes) and the lowest engineering cost, which aligns with the company's priorities of speed and cost efficiency, even though its MAE (1.5) is slightly worse.
- D. Incorrect.
This option is incorrect because the linear regression model meets the company's requirements for faster training time and lower engineering costs, even if it does not provide the lowest error.