networkx-social

NetworkX Graph Analysis

Graph analysis, community detection, centrality measures, knowledge graphs, bipartite networks, and visualization.

Graph Creation

import networkx as nx

G = nx.Graph()
G.add_edge('Alice', 'Bob', weight=3)
G.add_edges_from([('Bob', 'Carol'), ('Carol', 'Dave'), ('Alice', 'Dave')])

D = nx.DiGraph()                          # directed graph
G = nx.read_edgelist('network.txt')       # from edge list file
G = nx.from_pandas_edgelist(df, 'source', 'target')  # from DataFrame

Knowledge Graph Construction

# Build a knowledge graph from (subject, predicate, object) triples
KG = nx.DiGraph()
triples = [
    ("Python", "is_a", "Language"), ("Pandas", "depends_on", "Python"),
    ("NumPy", "depends_on", "Python"), ("Pandas", "depends_on", "NumPy"),
]
for subj, pred, obj in triples:
    KG.add_edge(subj, obj, relation=pred)

# Query: all dependencies of Pandas
deps = list(nx.descendants(KG, "Pandas"))

# Subgraph around a node (ego graph)
ego = nx.ego_graph(KG, "Python", radius=2, undirected=True)

Centrality Measures

dc = nx.degree_centrality(G)              # fraction of connected nodes
bc = nx.betweenness_centrality(G)         # shortest-path intermediary
cc = nx.closeness_centrality(G)           # inverse avg distance
ec = nx.eigenvector_centrality(G, max_iter=1000)  # neighbor importance
pr = nx.pagerank(D, alpha=0.85)           # PageRank (directed)

for node, score in sorted(bc.items(), key=lambda x: -x[1])[:5]:
    print(f"{node}: {score:.4f}")

Community Detection

from networkx.algorithms.community import louvain_communities, modularity
from networkx.algorithms.community import label_propagation_communities, greedy_modularity_communities

communities = louvain_communities(G, seed=42)        # modularity optimization
communities = list(label_propagation_communities(G))  # fast, non-deterministic
greedy = list(greedy_modularity_communities(G))       # greedy modularity

mod = modularity(G, communities)
print(f"Modularity: {mod:.4f}")