CompTIA DataX DY0-001 (V1) Practice Question

The correct answer is LASSO (Least Absolute Shrinkage and Selection Operator) regression. This model is the most suitable choice because it meets both of the scenario's requirements: handling multicollinearity and performing automatic feature selection. LASSO uses L1 regularization, which adds a penalty term proportional to the absolute value of the coefficients. A key feature of this penalty is its ability to shrink the coefficients of less important features to exactly zero, effectively removing them from the model. This results in a simpler, more interpretable (parsimonious) model that highlights the most significant predictors, which aligns perfectly with the business objective.

• Ridge regression is incorrect because while it effectively handles multicollinearity by using L2 regularization to shrink coefficients, it does not perform feature selection. The coefficients are shrunk towards zero but never become exactly zero, meaning all features are retained in the final model.

• Ordinary Least Squares (OLS) regression is incorrect because it performs poorly in the presence of strong multicollinearity. Multicollinearity violates one of the key assumptions of OLS, leading to unstable coefficient estimates with high variance, making the model unreliable for interpretation and prediction.

• Linear Discriminant Analysis (LDA) is incorrect because it is primarily a classification and dimensionality reduction technique, not a regression method designed for this type of feature selection problem. While LDA reduces dimensions by finding linear combinations of features that best separate classes, it does not provide a parsimonious regression model in the way that LASSO does.