Skip to content

Latest commit

 

History

History
32 lines (24 loc) · 2.38 KB

File metadata and controls

32 lines (24 loc) · 2.38 KB

Burst Metrics Methodology

Scope

This project uses standard statistical primitives and then composes them into app-specific burst windows, draft bins, and API payloads. The burstiness signal is the primary input to adaptive temporal scaling (see TEMPORAL_SCALING_CHARACTERIZATION.md), and density is used only for visualisation (density strip, count summary).

Metric Sources

Temporal burstiness

  • Goh, K.-I., & Barabasi, A.-L. (2008). Burstiness and memory in complex systems. EPL (Europhysics Letters), 81(4), 48002. https://doi.org/10.1209/0295-5075/81/48002
  • Source of the inter-event coefficient B = (sigma - mu) / (sigma + mu) used in the detect flow.

Entropy and spatial entropy

Jensen-Shannon divergence

What Is Custom

  • computeSpatialBBinned() supports selectable spatial formulas: ANN, entropy, JS divergence, and the balanced composite.
  • The balanced default combines entropy-based concentration with Jensen-Shannon divergence as clamp01(concentration * (0.25 + 0.75 * surprise)) so low-divergence bins still retain a stable floor.
  • combinedB in the API is a project-specific weighting of temporal and spatial scores.
  • Burst ranking, thresholds, and slice allocation are app logic, not borrowed directly from a single paper.

Code Map

  • src/components/dashboard-demo/lib/demo-burst-generation.ts - local draft-bin burst scoring.
  • src/lib/burst-detection.ts - shared burst scoring helpers and spatial metric.
  • src/app/api/adaptive/bursts/route.ts - API composition of temporal and spatial burst scores.

Summary

The temporal coefficient is literature-backed, the spatial ingredients are literature-backed, and the final burst pipeline is a project-specific composition of those pieces.