Google Professional Machine Learning Engineer Question 222
Single answerGoogle Cloud PlatformYou are tasked with building a machine learning model for a healthcare organization to predict patient readmission rates within 30 days of discharge. Due to regulatory requirements, the model must provide clear and interpretable explanations for its predictions. Which modeling technique would be the most appropriate to meet these interpretability requirements while maintaining reasonable accuracy?
- A
Linear Regression
- B
Random Forest
- C
Deep Neural Networks
- D
Gradient Boosted Trees
Show answer and explanation
Correct answer: A
Explanation
When interpretability is a key requirement, simpler models like Linear Regression are preferred because they provide transparent and straightforward explanations for predictions. Although other techniques such as Random Forest or Gradient Boosted Trees can offer higher accuracy, their interpretability is limited without additional tools, which may not meet strict regulatory demands in sensitive domains like healthcare.
- A. Correct.
Linear Regression is inherently interpretable since it provides clear coefficients for each feature, showing the contribution of each variable to the prediction. This makes it suitable for high interpretability requirements.
- B. Incorrect.
Random Forest models are less interpretable due to their ensemble nature, as they aggregate results from multiple decision trees. While feature importance can be extracted, it doesn't provide detailed interpretability for individual predictions.
- C. Incorrect.
Deep Neural Networks are highly complex and often considered black-box models, offering very limited interpretability without additional techniques like SHAP or LIME, which may not fully satisfy strict regulatory requirements.
- D. Incorrect.
Gradient Boosted Trees, like XGBoost or LightGBM, are more interpretable than deep learning methods but still lack the inherent simplicity of linear models. They require additional tools for explaining predictions, which adds complexity.