AI0-001 AI Models and Data Engineering Practice Question

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Gradient boosting machine

  Why it's wrong here

  Gradient boosting machines, such as XGBoost or LightGBM, are powerful ensemble methods that often achieve state-of-the-art performance on tabular data. However, they are also considered black-box models because the combination of many weak learners makes it challenging to explain individual predictions. While techniques like SHAP can provide post-hoc explanations, they add complexity and may not be as straightforward as logistic regression coefficients. In a scenario where interpretability is a primary requirement, gradient boosting is not the most appropriate choice.

• ✓

  Logistic regression

  Why this is correct

  Logistic regression is a linear model that provides coefficients for each feature, indicating the direction and magnitude of their influence on the predicted probability. This makes it highly interpretable, as stakeholders can understand how each feature contributes to the prediction. It is well-suited for binary classification tasks and performs reasonably well when the relationship between features and the log-odds is approximately linear. In this scenario, with 20 features and 10,000 samples, logistic regression can be trained efficiently and the coefficients can be explained to business users.

• ✗

  Random forest

  Why it's wrong here

  Random forest is an ensemble of decision trees that can capture complex non-linear relationships and often achieves high accuracy. However, it is less interpretable than a single decision tree or linear model because the ensemble combines many trees, making it difficult to explain individual predictions. While feature importance can be derived, it does not provide the same level of transparency as logistic regression coefficients. For a scenario requiring interpretability to business stakeholders, random forest would not be the best choice.

• ✗

  Support vector machine with RBF kernel

  Why it's wrong here

  Support vector machines with a radial basis function kernel are effective for non-linear classification but are generally not interpretable. The decision boundary is defined by support vectors in a high-dimensional space, and explaining predictions requires understanding the kernel transformation, which is not intuitive for business stakeholders. While SVMs can perform well, they are not suitable when interpretability is a key requirement. In this scenario, logistic regression would be preferred for its transparency.