Clustering Explorer — and whether your clusters are real

1

Load a dataset

⇪

Drop a CSV here, or click to browse

Numeric and categorical columns are scaled and encoded automatically

Or try:

2

Choose how many clusters

Columns used identifiers and constants are excluded automatically

Silhouette by k higher is better separated — the peak is a good starting point

Number of clusters (k) Algorithm

3

Are these clusters real?

Projection first two principal components — a 2-D shadow of a higher-dimensional space

Every algorithm, same data

4

What makes each cluster different

Not a table of averages — the features that sit furthest from the overall population, measured in standard deviations. That is what actually characterises a segment.