KOL Analytics
A graph analytics engine powered by real CMS Open Payments data for oncology specialists. Applies PageRank, betweenness centrality, and Louvain community detection to the top 150 KOLs by industry payment value — mapping influence networks, identifying community structure, and rendering an interactive force-directed visualization with live data from 61K+ oncology providers and 140K+ payment records.
Data Sources: CMS NPPES Provider Registry (61K+ oncology providers), CMS Open Payments 2021-2023 (140K+ payment records, 226 pharmaceutical payers). Graph analysis via PageRank centrality, betweenness centrality, and Louvain community detection. All data is real and fetched from Cloudflare D1.
Loading CMS Open Payments data & computing graph metrics...
PageRank computes global recursive influence using 50 iterations with 0.85 damping. Betweenness centrality is approximated via sampled BFS from 30 source nodes to maintain O(n*m) complexity. Composite scores blend PageRank (40%), degree (30%), and betweenness (30%).
Greedy modularity optimization inspired by the Louvain method. Iteratively moves nodes between communities to maximize Q, the ratio of within-community edge density to expected density under a null model. Converges in under 10 passes for networks of this scale.
Fruchterman-Reingold force-directed placement with 150 iterations and simulated annealing. Repulsive forces between all node pairs and attractive forces along edges produce spatial clustering that mirrors community structure. Seeded PRNG ensures deterministic output.