Exam AWS Certified Machine Learning - Specialty topic 1 question 168 discussion

B,C should be the answer

Apply principal component analysis (PCA) (Option B): PCA is a dimensionality reduction technique that transforms the original features into a smaller set of uncorrelated components, which can help in reducing the redundancy caused by highly correlated features. Remove a portion of highly correlated features from the dataset (Option C): By removing some of the highly correlated features, the data engineer can simplify the model and reduce multicollinearity, which can improve the model’s performance and interpretability

BD should be the answer

PCA is sensitive to the variance of features, so it's a common practice to standardize (e.g., z-score normalization) or scale (e.g., min-max scaling) the features before applying PCA. If the features are on different scales, it can distort which principal components are viewed as the most important.

Well, first time that I go with the Suggested Answer. D - B is the way: We want to solve the base correlation problem. That said, Letters A - E don't solve this problem, so they're wrong. Letter C partially solves the problem, so it is wrong. As we want steps, the correct alternatives are: D (ensure that all variables are on the same scale) and B (apply PCA that removes all correlation from the base while keeping most of the information). Again, from my perspective (C) is vague and using (B) removes the necessity of drop highly correlated features.

the most effective steps to address the issue of high correlation among the features in the dataset are removing a portion of highly correlated features and applying principal component analysis (PCA) for dimensionality reduction. These steps will help improve the data quality and predictive performance of the model.

PCA is a widely used technique for reducing the dimensionality of high-dimensional datasets while retaining as much of the original variability as possible. It is particularly useful when dealing with highly correlated features. Removing a portion of highly correlated features can be another effective way to address the issue of high correlation. By removing some of the correlated features, the model can become less complex and less prone to overfitting.

MinMax scaling does nothing to fix the issue here

BC for me, minmax scaler cannot remove multicollinearity?