AWS Certified Machine Learning – SpecialtyExploratory Data AnalysisHard
A machine learning engineer is tasked with building a model to predict equipment failure based on sensor readings. The dataset contains 100 highly correlated sensor features, making the model complex and prone to overfitting. They need to reduce the dimensionality of the dataset while retaining as much variance as possible for the prediction task. Which technique is most appropriate?
- APrincipal Component Analysis (PCA)
- BFeature selection using Lasso regularization
- COne-Hot Encoding
- DStandardization
Show answer & explanationAnswer & explanation
Correct answer: A. Principal Component Analysis (PCA)
Principal Component Analysis (PCA) is a dimensionality reduction technique that transforms a set of correlated variables into a smaller set of uncorrelated variables called principal components. It achieves this by finding directions (components) that capture the maximum variance in the data, making it ideal for reducing dimensionality while preserving information in highly correlated features.
Why the other options are wrong
- B. Lasso regularization is a feature selection technique that sets some feature coefficients to zero, effectively removing them. While it reduces dimensionality, it doesn't create new, uncorrelated features that capture overall variance like PCA.
- C. One-Hot Encoding is used to convert categorical features into numerical format and increases dimensionality, which is the opposite of the goal.
- D. Standardization scales features to a common range/distribution but does not reduce the number of features or address multicollinearity.
Principal Component Analysis (PCA)
A statistical procedure that uses an orthogonal transformation to convert a set of observations of possibly correlated variables into a set of linearly uncorrelated variables called principal components.
- Dimensionality reduction technique.
- Captures maximum variance in fewer dimensions.
- Useful for handling multicollinearity and preventing overfitting.
Memory trick: PCA: 'P'owerful 'C'ompression, 'A'll variance preserved.