A data engineer finds many of 100 features in a customer tabular dataset are highly correlated. Which actions should they take to address multicollinearity? (Choose two.)

Multicollinearity occurs when independent variables in a regression model are highly correlated, which can lead to unstable and unreliable coefficient estimates. Principal Component Analysis (PCA) addresses this by transforming the correlated features into a smaller set of uncorrelated principal components, effectively reducing dimensionality while retaining most of the variance. Alternatively, removing some of the highly correlated features directly from the dataset is a straightforward way to eliminate the redundancy and reduce multicollinearity. Training with a linear-based algorithm does not address multicollinearity; in fact, it can be negatively impacted by it. Min-max feature scaling normalizes data but does not reduce correlation between features. One-hot encoding is for converting categorical variables into a numerical format and is unrelated to multicollinearity among existing numerical features.