Module multivariate 

chronix-multivariate

Multi-series analysis for the Chronix time-series database.

Capabilities

Feature	Key Types
Series alignment	MultiSeriesContext, ColumnarMatrix
Correlation	PearsonCorrelation, SpearmanCorrelation, KendallTau
Rolling / lag correlation	RollingCorrelation, LagCorrelation
Correlation matrix	CrossCorrelationMatrix
Anomaly detection	MahalanobisDetector, IsolationForestDetector, PcaAnomalyDetector
Forecasting	MultiLinearRegression, VarModel
Derived series	ArithmeticExpr, DerivedSeriesEngine
Composite signals	CompositeSignalEngine, CompositeSignalRule

Overview

MultiSeriesContext aligns multiple time-series to a common time grid (via interpolation or forward-fill), producing a ColumnarMatrix that feeds into correlation, anomaly detection, and forecasting APIs.

All multivariate anomaly detectors implement MultivariateAnomalyDetector and produce MultivariateAnomalyScore for uniform consumption.

CompositeSignalEngine evaluates boolean rules over multiple series and emits signals via an async channel when conditions are met.

Structs

AnalyticsResults

Results from upstream analytics (forecasts, anomaly scores) passed to composite signal conditions.

ArithmeticExpr

Arithmetic on two named series.

ColumnarMatrix

Column-major matrix for aligned multi-series data.

CompositeSignal

A composite signal emitted when conditions are met.

CompositeSignalEngine

Engine that evaluates composite signal rules.

CompositeSignalRule

A rule that evaluates against a multi-series context.

CrossCorrelationMatrix

Full N×N pairwise Pearson correlation matrix.

DerivedSeriesDefinition

Definition of a derived series.

DerivedSeriesEngine

Engine that evaluates derived series in topological order.

GrangerCausalityResult

Result of a pairwise Granger causality test.

IsolationForestDetector

Isolation Forest multivariate anomaly detector.

KendallTau

Kendall Tau-a rank correlation.

LagCorrelation

Lag correlation: Pearson at various time offsets.

LazyDerivedSeries

Lazy wrapper around a derived series definition that caches its result.

LogReturnExpr

Log returns: ln(p_t / p_{t-1}).

MahalanobisDetector

Multivariate anomaly detection via Mahalanobis distance.

MultiLinearRegression

OLS regression: multiple predictors → single target.

MultiSeriesContext

Time-aligned context for multi-series analysis.

MultivariateAnomalyScore

Score for a multivariate anomaly at a given timestamp.

MultivariateForecastResult

Result of a multivariate forecast.

PcaAnomalyDetector

PCA-based multivariate anomaly detection.

PearsonCorrelation

Pearson correlation coefficient.

RollingCorrelation

Rolling Pearson correlation with O(1) update per point.

RollingStatExpr

Rolling statistic expression.

SignalReceiver

Receiver end of the signal delivery channel.

SignalSender

A channel-based signal delivery layer.

SpearmanCorrelation

Spearman rank correlation.

VarModel

After fitting, the model verifies that the VAR companion matrix has a spectral radius below stability_threshold (default: 1.05). Models exceeding this are rejected as unstable. The QR-based eigenvalue computation correctly captures complex conjugate eigenvalue pairs (oscillatory dynamics) that power iteration would miss. Callers can override the threshold via new_with_stability_threshold().

Enums

ArithOp

Binary arithmetic operators for derived series.

MultivariateError

Errors produced by multivariate analysis operations.

RollingStat

Supported rolling statistics.

Constants

DEFAULT_SIGNAL_CHANNEL_CAPACITY

Default channel capacity for signal delivery.

Traits

CorrelationMethod

Common trait for pairwise correlation methods.

DerivedSeriesExpr

An expression that produces a derived series from a context.

MultivariateAnomalyDetector

Multivariate anomaly detection contract.

MultivariateForecastModel

Multivariate forecast model contract.

Functions

signal_channel

Creates a new bounded signal delivery channel (sender + receiver).