Google Professional Machine Learning Engineer Question 223
Single answerGoogle Cloud PlatformYou are developing a machine learning model for a healthcare application that predicts the likelihood of a patient developing a certain condition. The healthcare professionals who will use the model require clear, interpretable explanations for each prediction to comply with medical regulations and ensure trust in the system. Which modeling approach should you choose to best meet the interpretability requirements?
- A
Linear Regression or Logistic Regression with feature importance analysis
- B
Deep Neural Networks with Layer-wise Relevance Propagation (LRP)
- C
Random Forest with SHAP (SHapley Additive exPlanations) values
- D
Support Vector Machines with a Radial Basis Function (RBF) kernel
Show answer and explanation
Correct answer: A
Explanation
When interpretability is a critical requirement, simpler models like Linear Regression or Logistic Regression are preferred because they are inherently interpretable. The coefficients of these models provide a direct and clear explanation of the relationship between input features and the target variable, which is especially important in high-stakes applications like healthcare. While advanced techniques such as SHAP or LRP can be used to explain complex models, they add unnecessary complexity when simpler methods suffice.
- A. Correct.
Linear Regression and Logistic Regression are inherently interpretable models. Coefficients directly indicate the relationship between features and the target variable, making them suitable for scenarios where interpretability is critical.
- B. Incorrect.
Deep Neural Networks are complex models that are difficult to interpret natively. While Layer-wise Relevance Propagation (LRP) can help explain predictions, it is not as inherently interpretable as simpler models like Linear Regression.
- C. Incorrect.
Random Forests are not inherently interpretable, although SHAP values provide post-hoc explanations. However, this approach is more complex than necessary for the stated requirement of clear and straightforward interpretability.
- D. Incorrect.
Support Vector Machines with an RBF kernel are difficult to interpret due to their non-linear decision boundaries and lack of native feature importance metrics, making them unsuitable for this use case.