Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/eugene-kamenev/tsmp4j
Time Series with Matrix Profile in Java
https://github.com/eugene-kamenev/tsmp4j
algorithms anomaly-detection clustering data-mining knn matrixprofile motif-analysis motif-discovery similarity-search timeseries timeseries-analysis
Last synced: about 2 months ago
JSON representation
Time Series with Matrix Profile in Java
- Host: GitHub
- URL: https://github.com/eugene-kamenev/tsmp4j
- Owner: eugene-kamenev
- License: apache-2.0
- Created: 2023-08-01T09:18:23.000Z (over 1 year ago)
- Default Branch: master
- Last Pushed: 2024-08-26T06:46:37.000Z (4 months ago)
- Last Synced: 2024-08-26T08:25:22.805Z (4 months ago)
- Topics: algorithms, anomaly-detection, clustering, data-mining, knn, matrixprofile, motif-analysis, motif-discovery, similarity-search, timeseries, timeseries-analysis
- Language: Java
- Homepage:
- Size: 432 KB
- Stars: 4
- Watchers: 4
- Forks: 1
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
## Time Series with Matrix Profile in Java
This repository contains Matrix Profile algorithms implemented in Java.
It is an attempt to port algorithms presented
in [tsmp](https://github.com/matrix-profile-foundation/tsmp).
[The Matrix Profile](https://www.cs.ucr.edu/~eamonn/MatrixProfile.html), has the potential to
revolutionize time series data mining because of its generality,
versatility, simplicity and scalability.
In particular, it has implications for time series motif discovery, time series joins, shapelet
discovery (classification), density estimation, semantic segmentation, visualization, rule
discovery, clustering etc.
This library includes the following algorithms to compute matrix profile:
Z-Normalized Euclidean Distance:
1. [STAMP](/src/main/java/com/github/eugene/kamenev/tsmp4j/algo/mp/stamp/STAMP.java) - Anytime matrix profile algorithm
2. [STOMP](/src/main/java/com/github/eugene/kamenev/tsmp4j/algo/mp/stomp/STOMP.java) - Scalable ordered matrix profile algorithm
3. [STOMPI](/src/main/java/com/github/eugene/kamenev/tsmp4j/algo/mp/stompi/STOMPI.java) - Incremental matrix profile algorithm
4. [SKIMP](/src/main/java/com/github/eugene/kamenev/tsmp4j/algo/pmp/SKIMP.java) - Pan matrix profile algorithm
5. [MPX](/src/main/java/com/github/eugene/kamenev/tsmp4j/algo/mp/mpx/MPX.java) - Matrix profile algorithm not based on FFT
6. [MP-DIST (MASS2)](/src/main/java/com/github/eugene/kamenev/tsmp4j/algo/mp/mass/MASS2.java) - Fast distance search algorithm based on FFT
7. [ContrastProfile](/src/main/java/com/github/eugene/kamenev/tsmp4j/algo/cp/ContrastProfileAlgorithm.java)
8. [PanContrastProfile](/src/main/java/com/github/eugene/kamenev/tsmp4j/algo/cp/PanContrastProfileAlgorithm.java)
9. [RelativeFrequencyMatrixProfile](/src/main/java/com/github/eugene/kamenev/tsmp4j/algo/cp/RelativeFrequencyMatrixProfileAlgorithm.java)
10. [RelativeFrequencyContrastProfile](/src/main/java/com/github/eugene/kamenev/tsmp4j/algo/cp/RelativeFrequencyContrastProfileAlgorithm.java)
11. [FLUSS](/src/main/java/com/github/eugene/kamenev/tsmp4j/algo/fluss/FLUSS.java) - timeseries segmentation based on matrix profile
Pure Euclidean Distance:
1. [AAMP](/src/main/java/com/github/eugene/kamenev/tsmp4j/algo/mp/aamp/AAMP.java)
Additionally, this library includes extra algorithms not related to matrix profile:
1. [MWF](/src/main/java/com/github/eugene/kamenev/tsmp4j/algo/extras/windowfinder/MWF.java) - Domain agnostic window size finder
2. [trendSegmentR](/src/main/java/com/github/eugene/kamenev/tsmp4j/algo/extras/tguw/Trend.java) - Detection of linear trend changes for univariate time series
3. [TGUW](/src/main/java/com/github/eugene/kamenev/tsmp4j/algo/extras/tguw/TGUW.java) - Tail-Greedy Unbalance Haar Wavelet decomposition
More algorithms will be added in the future.
## Usage
All algorithms for matrix profile are built around [RollingWindowStatistics](/src/main/java/com/github/eugene/kamenev/tsmp4j/stats/RollingWindowStatistics.java) object.
It simply computes statistics required to run Matrix Profile algorithms on the fly, in a circular buffer manner, hence allows streaming data processing out of the box.
### Streaming case
```java
DoubleStream stream = ... // your data stream
var bs = 1024; // statistics buffer size
var w = 10; // window for MP algorithm
var stamp = new STAMP(w, bs);
stream.forEach(stamp::update); // it keeps statistics updated, not the matrix profile
var matrixProfile = stamp.get(); // execute MP algorithm for statistics collected
```
### Single batch case
```java
double[] data = ... // your data
var w = 10; // window for MP algorithm
var matrixProfile = STAMP.of(data, w);
```
Please refer to [tests](/src/test/groovy/com/github/eugene/kamenev/tsmp4j/algo) for more examples.