Pattern-based change detection made easy with jKarma

Quickly define and execute custom unsupervised change detectors for evolving data with few lines of Java code.

jKarma allows the modular definition and easy execution of Pattern-Based Change Detection (PBCD) algorithms on evolving data with reduced implementation efforts. It is completely independent from third parties data sources and analytics frameworks. Every PBCD algorithm built with jKarma is:


Changes are sought by monitoring a symbolic model (relational patterns) learned from data.


Changing data are discriminated from ordinary ones without the need of a training phase on finely tuned datasets


Changes are sought without making any prior assumption on the data distribution.


PBCDs decide about the presence of a change, also by explaining them with descriptive pattern-based models.


jKarma is fully documented. You can check the developer documentation and the associated scientific publications.

Intro to Pattern-based Change Detection

A brief introduction to the change detection problem solved with pattern-based approaches.

The jKarma Tutorial

An tutorial showing how to define and execute PBCD algorithms on evolving data with jKarma.

Demonstrative Projects

Public repository containing different demonstrative projects using jKarma.


The reference javadoc for Java developers.

Scientific Publications

Angelo Impedovo, Corrado Loglisci, Michelangelo Ceci, Donato Malerba. jKarma: a Highly-Modular Framework for Pattern-based Change Detection on Evolving Data. Knowledge-Based Systems, accepted 2019, in press.

Corrado Loglisci, Michelangelo Ceci, Angelo Impedovo, Donato Malerba. Mining Microscopic and Macroscopic Changes in Network Data Streams. Knowledge-Based Systems 161: 294-312 (2018)

Corrado Loglisci, Angelo Impedovo, Michelangelo Ceci, Donato Malerba. Mining Microscopic and Macroscopic Changes in Network Data Streams (Discussion Paper). SEBD 2019

Angelo Impedovo, Michelangelo Ceci, Toon Calders. Efficient and Accurate Pattern-based Change Detection in Dynamic Networks. Discovery Science 2019

Angelo Impedovo, Corrado Loglisci, Michelangelo Ceci. Temporal Pattern Mining from Evolving Networks. CoRR abs/1709.06772 (2017), presented at PhD Forum@ECML-PKDD 2017