Note
Click here to download the full example code
Communities in Zachary's karate club¶
Run pymocd.rimpso on Zachary's karate club, draw the
partition it returns, and score it against the club's recorded split.
Detect communities¶
Every detector takes a NetworkX graph with integer node ids (or an igraph
graph) and returns a dict mapping each node to a community id.
import matplotlib.pyplot as plt
import networkx as nx
import pymocd
G = nx.karate_club_graph()
partition = pymocd.rimpso(G)
print(f"{G.number_of_nodes()} nodes, {len(set(partition.values()))} communities")
Out:
Draw the partition¶
One colour per community. Isolated nodes carry community -1 and would be
drawn as a group of their own; drop them from the dict first if you do not
want them shown.
pos = nx.spring_layout(G, seed=7)
communities = sorted(set(partition.values()))
cmap = plt.colormaps["tab10"] if len(communities) <= 10 else plt.colormaps["tab20"].resampled(len(communities))
fig, ax = plt.subplots(figsize=(6, 4.5))
nx.draw_networkx_edges(G, pos, ax=ax, alpha=0.25)
for i, c in enumerate(communities):
nodes = [v for v in G if partition[v] == c]
nx.draw_networkx_nodes(G, pos, nodelist=nodes, node_color=[cmap(i)], node_size=180, ax=ax, label=f"community {c}")
nx.draw_networkx_labels(G, pos, font_size=7, ax=ax)
ax.legend(loc="lower left", fontsize=8, frameon=False)
ax.set_axis_off()
fig.tight_layout()
plt.show()
Score against ground truth¶
NetworkX records which faction each member joined after the split in the
club node attribute. pymocd.gt_metrics returns NMI,
AMI, ARI and pairwise F1 in one call.
gt = {v: int(G.nodes[v]["club"] != "Mr. Hi") for v in G}
nmi, ami, ari, f1 = pymocd.gt_metrics(partition, gt)
print(f"NMI={nmi:.3f} AMI={ami:.3f} ARI={ari:.3f} F1={f1:.3f}")
Out:
Total running time of the script: ( 0 minutes 0.089 seconds)
Download Python source code: plot_karate_club.py