Principal Components Analysis

What is Principal Component Analysis?

Principal Component Analysis (PCA) is a statistical procedure that uses an orthogonal transformation to convert a set of observations of possibly correlated variables into a set of values of linearly uncorrelated variables called principal components. This transformation is defined in such a way that the first principal component has the largest possible variance (that is, accounts for as much of the variability in the data as possible), and each succeeding component, in turn, has the highest variance possible under the constraint that it is orthogonal to the preceding components. The resulting vectors (each being a principal component) are an uncorrelated orthogonal basis set.

Why Use PCA?

PCA is used predominantly to reduce the dimensionality of a data set while retaining as much information as possible. This is achieved by keeping the principal components with the largest variance and ignoring the lower variance components, which are assumed to contain the noise of the dataset. PCA is sensitive to the relative scaling of the original variables; hence, data normalization is a crucial preprocessing step.

Applications of PCA include but are not limited to:

  • Quantitative finance: for risk management and portfolio optimization.
  • Image processing: for facial recognition and image compression.

  • Genomics: for reducing the dimensionality of genetic data.
  • Signal processing: for signal de-noising and data compression.

How PCA Works

The steps to perform PCA include:

  1. Standardizing the data: PCA is affected by scale, so the data needs to be normalized.
  2. Calculating the covariance matrix: To understand how the variables of the input data are varying from the mean with respect to each other.
  3. Computing the eigenvectors and eigenvalues of the covariance matrix: To identify the principal components.

  4. Choosing components and forming a feature vector: By ranking the eigenvalues in descending order and choosing the top k eigenvectors.
  5. Deriving the new data set: This is done by multiplying the original data set by the feature vector.

Understanding Eigenvalues and Eigenvectors

In PCA, eigenvectors and eigenvalues of a covariance (or correlation) matrix represent the "core" of PCA: eigenvectors determine directions of the new feature space, and eigenvalues determine their magnitude. In other words, eigenvalues explain the variance of the data along the new feature axes.

Choosing the Number of Principal Components

The number of principal components retained in the analysis is a critical decision. In practice, the choice is often made based on the cumulative explained variance, which should be as high as possible, while keeping fewer components. A common rule of thumb is to keep the principal components that explain at least 85% of the variance.

Advantages and Disadvantages of PCA


  • Removal of multicollinearity: PCA helps in mitigating the problem of multicollinearity in the data by transforming the original variables into a new set of variables that are uncorrelated.
  • Reduction of overfitting: By reducing the dimensionality, PCA can help reduce the chances of overfitting in a predictive model.
  • Improvement in visualization: High-dimensional data can be difficult to visualize, but PCA can make this visualization easier by reducing the number of dimensions.


  • Interpretability: The principal components are linear combinations of the original variables and may not be easily interpretable.
  • Sensitivity to scaling: PCA is sensitive to the scaling of the variables, which means that the results can vary depending on how the data was scaled.
  • Data loss: While reducing dimensionality, some information is inevitably lost, which might be important depending on the context.

PCA in Practice

PCA is implemented in various programming languages and software, with functions readily available in libraries such as scikit-learn in Python. When using PCA, it's important to standardize the data first and decide on the number of components to retain based on the explained variance. The transformed data can then be used for further analysis, such as clustering or as input features for machine learning models.


PCA is a powerful tool for exploratory data analysis and preprocessing. It simplifies the complexity in high-dimensional data while retaining trends and patterns. However, it is essential to apply PCA correctly and interpret the results within the context of the data. When used appropriately, PCA can reveal the underlying structure of the data, reduce noise, and make other machine learning tasks more efficient and interpretable.


Jolliffe, I. T. (2002). Principal Component Analysis. Springer Series in Statistics.

Wold, S., Esbensen, K., & Geladi, P. (1987). Principal component analysis. Chemometrics and Intelligent Laboratory Systems, 2(1-3), 37-52.

Abdi, H., & Williams, L. J. (2010). Principal component analysis. Wiley Interdisciplinary Reviews: Computational Statistics, 2(4), 433-459.

Please sign up or login with your details

Forgot password? Click here to reset