Concept
PCA
PCA, also known as dimensionality reduction, is a Machine Learning concept. The Library holds 1 implementations, each one a working definition you can pull into Quant.
Top PCA indicators
1 total
What is PCA?
Principal component analysis (PCA) re-expresses a set of correlated variables as a smaller set of uncorrelated ones. Technically it finds the eigenvectors of the data's covariance (or correlation) matrix: each eigenvector is a principal component, a weighted combination of the original variables, and the components are ordered by how much of the total variance they explain. Keeping only the first few compresses the data while retaining as much of the total variance as any linear projection of that dimension can, which is why PCA is the canonical dimensionality-reduction technique.
Markets give it two natural jobs. Run across a basket of assets, the first component usually captures the shared move (loosely, the market factor), while later components pick up rotation and spread effects; each asset's loading measures how strongly it participates. Run across an indicator stack, PCA collapses dozens of overlapping inputs into a few orthogonal factors, a common feature-engineering step before model fitting, since redundant inputs inflate overfitting risk without adding information.
How traders use it
- Extracting a market factor: the first principal component of basket returns typically behaves like an index, and an asset that departs from what its loading implies becomes a relative-strength or pairs-trading candidate.
- Compressing features: replacing many correlated oscillators and averages with two or three components that carry most of the variance, so a downstream model estimates fewer parameters from the same amount of data.
- Reading concentration: when the first component's share of explained variance rises, assets are moving as one block (risk-on/risk-off conditions); when it falls, dispersion is increasing and cross-sectional selection matters more.
Related concepts · Features & pipeline
Concept family
Machine Learning
32 concepts mapped · 21 in the Library
PCA FAQ
Do you need to standardize data before running PCA?
Usually, yes. PCA is scale-sensitive: variables with larger numeric ranges dominate the covariance matrix regardless of structure. Z-scoring each input, or equivalently working from the correlation matrix, puts variables on comparable footing. Skipping this is defensible only when the inputs already share units and magnitudes, such as daily returns across broadly similar assets.
What do principal components actually mean in trading terms?
They are statistical constructs, not guaranteed economic factors. In practice the first component of an asset basket usually tracks the common market move, and in yield curves the first three are conventionally read as level, slope, and curvature. But component composition can shift as the estimation window rolls, so interpretations should be re-checked rather than assumed permanent.
Build PCA your way.
Quant writes, tests, and refines it with you — then it runs on LuxAlgo charting or ports to TradingView.