Clustering Specialist — Compare Workflow

Description: Evaluate multiple approaches or versions

When to Use

Use the compare workflow when you need to evaluate multiple approaches or versions.

Input Requirements

• Datasets with feature scaling specifications and distance metric requirements
• Domain knowledge about expected segment structure and count ranges
• Cluster count determination criteria (elbow, silhouette, gap statistic)
• Downstream use case requirements for cluster-based segmentation

Process

1. Initialize — Set up the compare context for Clustering Specialist
2. Execute — Perform the compare operation following Clustering Specialist's style
3. Validate — Check output against quality gates
4. Handoff — Deliver results to downstream personas

Output

• Trained clustering models with optimal cluster count justification
• Cluster quality reports with silhouette, CH, and DB index scores
• Cluster stability analysis with bootstrap confidence intervals
• Cluster profile summaries with centroid descriptions and segment characteristics

Quality Gates

• Cluster count must be justified using multiple determination methods
• Cluster quality must be evaluated with silhouette, Calinski-Harabasz, and Davies-Bouldin
• Cluster stability must be assessed through bootstrapped resampling
• Feature scaling decisions must be documented with impact analysis