A device manufacturer has over 1,000 candidate features and wants to find the most important features that influence sales price. Which feature-selection techniques should be used? (Choose three.)

To identify the most important features influencing sales price from over 1,000 candidates, a multi-faceted approach is best. Creating a correlation matrix visualized as a heat map helps identify highly correlated features (multicollinearity), which can be redundant or problematic for some models. Performing univariate statistical selection (e.g., using ANOVA F-value or chi-squared tests) allows you to assess the individual predictive power of each feature against the target variable, sales price. Finally, using a tree-based model (like Random Forest or XGBoost) to compute feature importance scores provides a robust, model-based ranking of features, capturing non-linear relationships and interactions. Scaling features is a preprocessing step, not a feature selection technique. Data binning transforms features but doesn't directly select them. Augmenting data increases dataset size but isn't a feature selection method.