Online PCA Analyzer

The Online PCA Analyzer accepts numeric data matrix (samples × features) and returns principal axes, scores, and variance metrics. It standardizes data by default to zero mean and unit variance, then computes the covariance matrix and its eigenvectors, or uses SVD for numerical stability.
Core outputs include the components (loadings), the transformed scores for each sample, and explained variance per component along with cumulative variance. Users can select the target number of components k (1 ≤ k ≤ min(n_features, n_samples)) to reduce dimensionality while preserving most of the information.
The tool supports data-preparation practices such as handling missing values, optional scaling methods, and validation of inputs. It is suitable for researchers performing exploratory data analysis, feature engineering, or pre-processing for machine learning pipelines.
Unique value propositions include clear variance decomposition, compatibility with SVD-based PCA, and outputs suitable for downstream tasks like clustering or visualization. Use cases include gene expression reduction, image feature compression, and market data simplification.