PCA – Principal Component Analysis
Overview
Principal Component Analysis (PCA) reduces
high‑dimensional data into fewer components while retaining most variance. It
helps simplify data, remove noise, and prepare it for further modeling.
Where to Find PCA
Pipeline → Model Studio → Data Preparation
→ PCA
Before You Start
• Dataset must contain continuous numerical
columns.
• Standardization recommended for accurate PCA.
• Null values in selected columns must be handled.
• PCA is an end‑task for lineage (lineage stops here).
Configuring PCA
• Features – Select numerical columns.
|
Field |
Result |
Remarks |
|
Independent Variables |
Columns selected for PCA input. |
Must be numeric; no null values; multi-select. |
|
Copy |
Whether data is copied before processing. |
True recommended; False modifies data in place. |
|
Whiten |
Scales components to unit variance. |
May reduce interpretability. |
|
SVD Solver |
Method for Singular Value Decomposition. |
auto, full, arpack, randomized. |
|
Tolerance |
Threshold for convergence. |
Applies mainly to arpack solver. |
|
Iterated Power |
Iterations to improve randomized SVD accuracy. |
Used when solver=randomized. |
|
N Oversamples |
Extra vectors for approximation. |
Default 10; used for randomized solver. |
Viewing Results
Explore view shows:
• Eigenvalues
• Eigenvectors
• Proportion of Variance
• Component loadings
Notes
• PCA does not track lineage forward.
• Works with import/export, save‑as, and metadata refresh.
• Compatible with Python, RubiNotebook, RubiSQL, RubiSpark.