pathpy is an OpenSource python package for the analysis of time series data on networks using higher-order and multi-order graphical models.
The package is specifically tailored to analyze temporal networks as well as sequential data that capture multiple short, independent paths observed in an underlying graph topology.
Examples for data that can be analysed with pathpy include time-stamped social networks, user click streams in information networks, biological pathways, or traces of information propagating in social media.
Unifying the analysis of pathways and temporal networks, pathpy provides various methods to extract time-respecting paths from time-stamped network data.
It extends (and will eventually supersede) the package pyTempnets.
pathpy facilitates the analysis of temporal correlations in time series data on networks.
It uses a principled model selection technique to infer higher-order graphical representations that capture both topological and temporal characteristics.
It specifically allows to answer the question when a network abstraction of time series data is justified and when higher-order network representations are needed.
The theoretical foundation of this package, higher-order network models, was developed in the following research works:
- I Scholtes: When is a network a network? Multi-Order Graphical Model Selection in Pathways and Temporal Networks, In KDD'17 - Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Halifax, Nova Scotia, Canada, August 13-17, 2017
- I Scholtes, N Wider, A Garas: Higher-Order Aggregate Networks in the Analysis of Temporal Networks: Path structures and centralities, The European Physical Journal B, 89:61, March 2016
- I Scholtes, N Wider, R Pfitzner, A Garas, CJ Tessone, F Schweitzer: Causality-driven slow-down and speed-up of diffusion in non-Markovian temporal networks, Nature Communications, 5, September 2014
- R Pfitzner, I Scholtes, A Garas, CJ Tessone, F Schweitzer: Betweenness preference: Quantifying correlations in the topological dynamics of temporal networks, Phys Rev Lett, 110(19), 198701, May 2013
pathpy extends this approach towards multi-layer graphical models that capture temporal correlations at multiple length scales simultaneously.
An illustrative example for a collection of paths (left) and a multi-order graphical representation is shown below.
All mathematical details of the framework can be found in this recent research paper.
pathpy is pure python code. It has no platform-specific dependencies and should thus work on all platforms.
It builds on numpy and scipy.
The latest version of pathpy can be installed by typing:
> pip install git+git://github.com/IngoScholtes/pathpy.git
A comprehensive educational tutorial which shows how you can use pathpy to analyze data on pathways and temporal networks is available online.
Moreover, a tutorial which illustrates the abstraction of higher-order networks in the modeling of dynamical processes in temporal networks is available here.
The latter tutorial is based on the predecessor library pyTempNets. Most of its features have been ported to pathpy.
The code is fully documented via docstrings which are accessible through python's built-in help system. Just type help(SYMBOL_NAME) to see the documentation of a class or method. A reference manual is available here.
The first public beta release of pathpy (released February 17 2017) is v1.0-beta. Following versions are named MAJOR.MINOR.PATCH according to semantic versioning. The date of each release is encoded in the PATCH version.
The research behind this data analysis framework was funded by the Swiss State Secretariat for Education, Research and Innovation (Grant C14.0036). The development of this package was generously supported by the MTEC Foundation in the context of the project The Influence of Interaction Patterns on Success in Socio-Technical Systems: From Theory to Practice.
Ingo Scholtes (project lead, development)
Luca Verginer (development, test suite integration)
Roman Cattaneo (development)
Nicolas Wider (testing)
pathpy is licensed under the GNU Affero General Public License.
(c) Copyright ETH Zürich, Chair of Systems Design, 2015-2017