Analysis Types¶

CommGraph provides several analysis algorithms for understanding communication patterns.

Centrality Analysis¶

Centrality measures identify the most important actors in a communication network.

PageRank¶

Google's PageRank algorithm adapted for email networks. Actors who receive messages from other important actors score higher.

commgraph analyze centrality --algorithm=pagerank --top=20

Best for: Identifying influential individuals who receive attention from other influential people.

Degree Centrality¶

Simple count of connections. Variants:

degree: Total connections (in + out)
in_degree: Incoming messages only
out_degree: Outgoing messages only

commgraph analyze centrality --algorithm=degree --top=20
commgraph analyze centrality --algorithm=in_degree --top=20
commgraph analyze centrality --algorithm=out_degree --top=20

Best for: Finding the most active communicators (degree) or most sought-after individuals (in_degree).

Betweenness Centrality¶

Measures how often an actor lies on the shortest path between other actors.

commgraph analyze centrality --algorithm=betweenness --top=20

Best for: Identifying information brokers and gatekeepers who control information flow.

Community Detection¶

Community detection identifies groups of actors who communicate more frequently with each other than with outsiders.

Louvain Algorithm¶

Fast, hierarchical community detection that optimizes modularity.

commgraph analyze community --algorithm=louvain

The --resolution parameter controls granularity:

Higher values (>1.0) produce more, smaller communities
Lower values (<1.0) produce fewer, larger communities

commgraph analyze community --algorithm=louvain --resolution=1.5

Label Propagation¶

Fast algorithm where nodes adopt the most common label among their neighbors.

commgraph analyze community --algorithm=label_propagation

Best for: Quick community detection on large graphs.

Bridge Detection¶

Bridge actors connect different communities and often serve as information gatekeepers.

commgraph analyze bridges --top=10

Output includes:

Actor identification
Communities they connect
Cross-community edge count
Betweenness score

Path Analysis¶

Analyze network paths between actors.

# Shortest path between two actors
commgraph analyze paths --from=alice@example.com --to=bob@example.com

# Network diameter and average path length
commgraph analyze paths --samples=100

Temporal Analysis¶

Detect patterns over time.

commgraph analyze temporal --window=24h

Identifies:

Communication bursts: Sudden spikes in activity
Trends: Increasing or decreasing communication over time
Patterns: Regular activity cycles

The --threshold parameter sets the z-score threshold for burst detection:

commgraph analyze temporal --window=24h --threshold=2.0

External Communication Analysis¶

Analyze communication patterns with external parties.

commgraph analyze external --top-domains=10

Output includes:

Top external domains by message count
Boundary spanners (internal actors with high external communication)
Inbound vs outbound ratios

Weight Profiles¶

All centrality analyses use weight profiles to adjust edge weights based on communication context:

Profile	Description	Use Case
`influence`	Higher weight for direct TO recipients	Who has organizational influence?
`information_flow`	Equal weight for all recipients	How does information spread?
`coordination`	Higher weight for CC/BCC recipients	Who coordinates activities?

commgraph analyze centrality --profile=influence
commgraph analyze centrality --profile=information_flow
commgraph analyze centrality --profile=coordination

See Weight Profiles for detailed configuration.

Output Formats¶

All analysis commands support multiple output formats:

# Table format (default)
commgraph analyze centrality --format=table

# JSON format
commgraph analyze centrality --format=json --output=results.json

# CSV format (centrality only)
commgraph analyze centrality --format=csv --output=results.csv