stumpy-dev/stumpy

Features

• Motif Discovery - Implements matrix profile algorithms to discover recurring patterns and motifs within long time series datasets.
• Matrix Profile Implementations - Implements matrix profile algorithms for calculating distance profiles to identify repeating patterns and discords.
• Pattern Identification - Identifies repeating patterns and anomalies by calculating the distance between every subsequence and its nearest neighbor.
• Time Series Anomaly Detection - Identifies discords and novelty patterns that deviate significantly from the rest of the time series data.
• Distributed Frameworks - Ships a scalable framework using Dask to parallelize heavy matrix profile computations across machines.
• Multidimensional Matrix Profiles - Computes matrix profiles for multidimensional data to identify similar patterns across different variables.
• Distributed Time Series Computation - Provides a scalable system to parallelize matrix profile computations across GPU hardware and Dask clusters.
• GPU-Accelerated Data Analysis - Offloads complex matrix calculations to GPU hardware to significantly reduce processing time for large datasets.
• Incremental Matrix Profiles - Calculates matrix profiles incrementally as new data arrives to monitor time series in real time.
• GPU-Accelerated Processing - Offloads complex matrix calculations to GPU hardware to reduce processing time for large datasets.
• Time Series Analysis Libraries - Provides a comprehensive Python library for analyzing time series data using matrix profile algorithms.
• Multidimensional Analysis - Implements multidimensional analysis to discover repeating patterns and motifs across multiple concurrent data streams.
• Multidimensional Motif Discovery - Provides capabilities to find repeating patterns across multiple concurrent data streams simultaneously.
• Streaming Analysis - Offers incremental computation to monitor time series in real time as new data points arrive.
• Distance Profile Computations - Calculates the distance between a specific query subsequence and every subsequence in a time series using the MASS algorithm.
• Matrix Profile Computation - Calculates distances between all subsequences of a time series using z-normalization and hardware acceleration.
• Chain Tracing - Discovers extended repeating patterns by computing anchored and unanchored chains of similar subsequences.
• Distributed Computing - Parallelizes distance profile calculations across multiple processors or servers to handle larger workloads.
• Dask Integrations - Distributes matrix profile computations across a cluster of machines using Dask integration.
• Chain Extraction - Analyzes sequential behavior by identifying temporally ordered sets of subsequence patterns.
• Time Series Segmenters - Detects regime changes and behavioral shifts by dividing time series into semantically distinct segments.
• Query Sequence Matching - Locates all occurrences of a specific query subsequence within a time series that fall below a defined distance threshold.
• Multidimensional Analyzers - Provides a framework for discovering motifs and patterns across multiple concurrent data streams simultaneously.
• Similarity Measurements - Computes the distance between two time series based on the overlap of their shared subsequences.
• Pan Matrix Profiles - Analyzes patterns of varying lengths by calculating matrix profiles across multiple different window sizes.
• Consensus Motif Identification - Identifies the most central conserved pattern shared across multiple different time series.
• Time Series Pattern Matching - Finds similar shapes by searching for specific motifs or conserved patterns across one or multiple time series.
• Time Series Analysis - Scalable library for modern time series matrix profile analysis.

Star history