Google Professional Machine Learning Engineer Question 219
Single answerGoogle Cloud PlatformYou are designing a machine learning model for a healthcare application that predicts whether a patient is at risk for a particular medical condition. The stakeholders have emphasized the need for high interpretability so that doctors can understand how predictions are made. What modeling approach should you prioritize?
- A
Use a linear regression model with feature weights for interpretability.
- B
Deploy a deep neural network with multiple hidden layers to achieve higher accuracy.
- C
Choose a decision tree model and visualize the decision paths for interpretability.
- D
Implement a random forest model and rely on feature importance metrics for interpretability.
Show answer and explanation
Correct answer: A
Explanation
Since the stakeholders explicitly require high interpretability, linear regression is the most appropriate choice in this scenario. Feature weights in linear regression are straightforward and easy to explain, making it suitable for environments like healthcare where understanding the reasoning behind predictions is critical. Other models like neural networks or random forests may offer higher accuracy but lack the level of interpretability required in this use case.
- A. Correct.
Linear regression provides high interpretability as feature weights directly indicate the contribution of each feature to the prediction. This aligns well with the stakeholder’s requirement to understand how predictions are made.
- B. Incorrect.
While deep neural networks can achieve high accuracy, they are generally considered black-box models and do not align with the requirement for high interpretability.
- C. Incorrect.
Decision tree models can provide some interpretability through visualization, but they may become overly complex for high-dimensional data, reducing their practical interpretability in some cases.
- D. Incorrect.
Random forest models can provide feature importance metrics, but the ensemble nature of the model makes it difficult to trace individual predictions back to specific feature values, which reduces their interpretability.