Google Professional Machine Learning Engineer Question 220
Select 2Google Cloud PlatformYou are designing a machine learning model for a healthcare application that predicts patient risks based on medical data. Due to regulatory requirements, the model must be interpretable to allow medical professionals to understand how predictions are made. Which of the following modeling techniques would best meet the interpretability requirements?
- A
Linear Regression
- B
Decision Trees
- C
Deep Neural Networks
- D
Random Forests
- E
SHAP (SHapley Additive exPlanations) applied to Gradient Boosted Trees
Show answer and explanation
Correct answers: A, B
Explanation
For applications with strict interpretability requirements, simple models like Linear Regression and Decision Trees are preferred because they provide straightforward explanations for their predictions. These models align with the need for transparency in regulated domains like healthcare, where understanding model predictions is critical. While techniques like SHAP can help interpret complex models, they add complexity and may not meet strict regulatory demands.
- A. Correct.
Linear Regression is highly interpretable because it provides direct coefficients that explain the relationship between input features and the target variable. This makes it suitable for use in regulated domains like healthcare.
- B. Correct.
Decision Trees are interpretable as they visually demonstrate decision paths and criteria used for predictions, making them easy for non-technical stakeholders to understand.
- C. Incorrect.
Deep Neural Networks are highly complex and often considered 'black-box' models due to the difficulty in interpreting how they make predictions. They are not suitable for scenarios requiring high interpretability.
- D. Incorrect.
Random Forests, while powerful, combine multiple decision trees, making them less interpretable than a single decision tree. Understanding the contribution of individual trees in a forest is challenging.
- E. Incorrect.
SHAP provides post-hoc interpretability for complex models like Gradient Boosted Trees, but it is not a modeling technique itself. Applying SHAP requires additional effort and may not fully satisfy strict interpretability requirements in regulated domains.