Part 1: EDA
- 1.1 Relational data as graphs
- basic igraph commands, main dataset, types of graphs
- 1.2 Exploratory data analysis (EDA)
- degree, paths, connected components
- 1.3 Node importance:
- coreness, centrality, betweenness
Part 2: Clustering
- 2.1 Transitivity
- triangles and other motifs, local and global clustering coefficients, cliques
- 2.2 Clustering: modularity, Louvain, Leiden, ECG and properties; other methods
- illustrations on various graphs; measures of cluster quality; ego-nets vs clusters
- 2.3 Random graph models and benchmarks (ex: Erdos-Renyi, ABCD)
- 2.4 Extra module: Anomaly detection
- illustration with football dataset
-
"Mining Complex Networks" (https://www.ryerson.ca/mining-complex-networks) by B. Kaminski, P. Pralat and F. Théberge.
-
Further notebooks and datasets to accompany this textbook can be found at: https://github.com/ftheberge/GraphMiningNotebooks
-
python-igraph documentation: https://igraph.readthedocs.io/en/0.10.2/
-
Introductory guide: https://towardsdatascience.com/newbies-guide-to-python-igraph-4e51689c35b4
The Python Notebook was created under the following conda environment:
conda create --name mining_graphs python=3.7.9 numpy pandas jupyter matplotlib pycairo scikit-learn
source activate mining_graphs
pip install igraph
pip install partition-igraph
python -m ipykernel install --user --name=mining_graphs ## add to Jupyter
The environment can be optionally created, managed and maintained reproducibly using make commands. To create and activate the conda environment:
make create_environment
source activate mining_graphs