Cluster Size Distribution
Number of agent-periods per cluster
Elbow + Silhouette Analysis
Optimal k selection metrics
Cluster Radar — Mean Metrics
Normalised mean feature values per cluster (0–1 scale)
PCA Scatter Plot — PC1 vs PC2
Each point is one agent-period. PC1 explains the most variance; PC2 the second-most.
Feature Distribution by Cluster
Box-like distribution per metric per cluster
Cluster Membership Over Time
Track how each agent's cluster assignment changes across periods
Agent Stability Rankings
Stability % = fraction of periods in modal cluster. Low values = frequent cluster switching.
Explained Variance per PC
Cumulative variance captured by each principal component
PCA Loadings Heatmap
Contribution of each original feature to each principal component
Documentation
Full documentation of the PCA + K-Means methodology, profiles, and drift analysis