Analysis#
Analysis is the operator’s catalog of algorithms and techniques for turning data into facts. Each entry below is a primitive, a well-understood method with known assumptions, known costs, and known failure modes. The operator composes them into larger systems: detectors that page on anomalies, dashboards that show trends, classifiers that route work, recommendations that surface priorities, signatures that match adversary behavior.
The aim of this section is to be a working reference an operator can read in one sitting and reach for during a build. Each page defines the technique, names the standard implementations, shows the operator-shaped use, and points at the next layer when the problem outgrows the primitive.
For mission-level analysis (target-centric tradecraft, OSINT synthesis, finished-intelligence production), see Analysis. For the engineering side of running analysis at scale (warehouses, streaming, ML pipelines), see Cloud and Databases. This section is the methods themselves.
Descriptive and inferential statistics. The substrate every later technique stands on.
Decomposition, smoothing, forecasting. For anything indexed by time.
Regression and classification. Given labels, predict the next one.
Clustering and dimensionality reduction. Find structure without labels.
Anomaly and change-point detection. Surface the unusual.
Shortest paths, centrality, community, link prediction.
Tokenisation, embedding, classification, extraction.
Online and approximate algorithms for unbounded data.
Information retrieval, ranking, similarity, recommendation. Vector space models and nearest-neighbor search.
Markov chains, HMMs, Kalman filters, MCMC. Stateful and uncertain processes.