Social, Knowledge, and Time
These features sit on the same /networks/ workspace. Switch graph mode (left) before generate, or use the timeline once a time field is mapped.
Social network analysis
Social mode replaces the standard centrality list with SNA-oriented jobs (POST /api/graphtheory/graphs/{id}/social-analysis):
| Option | Role |
|---|---|
| Influence score | Combined PageRank + betweenness style influence |
| Bridge score | Structural holes / brokerage |
| Reciprocity | Mutual directed ties |
| Clustering | Local clustering |
| Friend recommendations | Suggest links (count + min mutual friends) |
| Echo chambers | High in-group density vs out-group (threshold) |
| Influence paths | Bounded-length paths (max length, min strength) |
| Structural holes | Min bridge score filter |
Results appear in the social metrics table, not only as canvas colors.
Knowledge graphs
Knowledge mode treats vertices as typed entities and edges as relations.
- Load a built-in ontology (example: academic Person / Paper / Institution) or auto-generate from the current graph
- Validate schema against entity/relation types
- Apply inference (max iterations) for transitive / implied links
- Query related entities, semantic path, and custom knowledge queries
APIs live under /api/graphtheory/graphs/{id}/knowledge/*. This ontology is your domain graph, not the Runtime closed-loop catalog at /ontology/runtime.
Timeline and temporal algorithms
Map a time field and a bracket (minutes, hours, days, months, years). The canvas shows a histogram of buckets (server $dateTrunc; auto-coarser if the range would exceed 500 buckets).
Click a bucket (for example Jan 2023) to filter. From the timeline, run a temporal category:
| Category | Example algorithms |
|---|---|
| Structural shift | DeltaCon, spectral distance, BOCPD, PELT/CUSUM, top eigenspace |
| Influence rise/fall | PageRank over periods, k-core evolution, EWMA, rank-change tests |
| Community evolution | multilayer Leiden, label matching, event taxonomy, persistence/conductance |
| Tie formation / decay | link-prediction heuristics, gradient-boosted ranker, Cox PH, prediction evaluation |
| Cascade explosiveness | Hawkes, cascade metrics, changepoint |
Large corpora should send largeMode plus timeField and period bounds so the job does not stream the whole history into the browser.
Geo and analytics drawer
Optional settings on the saved network:
- Degree-distribution Chart.js widgets
- Natural-language charts via Workbench
/api/v1/analytics - Google Maps geo / earth charts (
VITE_GOOGLE_MAPS_API_KEY/env-config.js) when lat/lng fields (or geocoded address) exist
Writeback and export
After analysis:
| Writeback target | Effect |
|---|---|
| source | Merge metrics onto existing Mongo documents (_network_metrics prefix by default) |
| export | One score document per node in a target collection |
| full_export | Copy full source records plus scores |
Export on the toolbar downloads ecosystem-network-v1 JSON (mappings + settings). Import restores a model into network_configs.