MixedTrails

This repository contains the code used to produce the results from the following paper:

Martin Becker, Florian Lemmerich, Philipp Singer, Markus Strohmaier, Andreas Hotho
MixedTrails: Bayesian hypothesis comparison on heterogeneous sequential data
Data Mining and Knowledge Discovery, 2017; [link to publisher]

Presented at ECML PKDD 2017
Preprint available on ArXiv

The experiments are conducted on

a toy example about soccer strategies,
synthetic navigation paths on random graphs,
navigation from the game Wikispeedia, and
photowalking trails from the social photo-sharing platform Flickr.

Datasets

For the toy example and the synthetic experiments no additional data is needed.

Wikispeedia

The required Wikispeedia data can be downloaded from

https://snap.stanford.edu/data/wikispeedia.html

These two files are needed:

wikispeedia_paths-and-graph.tar.gz
wikispeedia_articles_plaintext.tar.gz

Directly extract both files into the wikispeedia folder. This should result in two folders:

wikispeedia_paths-and-graph
plaintext_articles

Flickr

For the Flickr data, please contact Martin Becker via e-mail:

becker@informatik.uni-wuerzburg.de

Run Experiments

For running the experiments you need python3. The requirements file for pip can be found in the root folder of this project (requirements.txt):

matplotlib==1.5.3
numpy==1.11.2
scipy==0.18.1
scikit_learn==0.18.1

After setting up your workspace, running the actual experiments is straight forward. For the toy example, the synthetic data as well as the Flickr experiments the corresponding Jupyter notebooks (exp-toy-offense.ipynb, exp-synthetic.ipynb, exp-flickr.ipynb respectively) are self contained. You can simply run them and have a look at the results.

For the Wikispeedia experiments you first need to run two scripts before the results can be visualized in the corresponding notebook (exp-wikispeedia-visualization.ipynb):

python exp-wikispeedia-prepare-data.py
python exp-wikispeedia.py

Contact