Character Clustering Lab

Load the character CSV, select what the computer can see, and compare clustering methods.

1. Load data

Choose character_data.csv. Everything runs locally in the browser.

2. Include features

Age / level Alignment Armed / unarmed Primary colour Human / nonhuman Image background colour Origin / location Original medium

3. Algorithm

Number of clusters5

Neighbourhood radius0.85

Minimum neighbours3

Show shared-feature links Show names

How to read it

Every visible line means two characters share at least one selected feature. Thicker/brighter lines mean more overlap. Node colour shows the cluster assigned by the selected algorithm.

Hover to see selected values and the five closest active characters according to Euclidean distance.

Load a CSV to begin.

Waiting for character data

Open the CSV using the control on the left.

Load your character_data.csv to create an interactive clustering map.

Mouse wheel: zoom | Drag background: pan | Drag node: move

Behind the scenes: from cards to clusters

The computer does not understand “USA”, “film”, or “good” as words. It encodes every category as on/off switches (one-hot encoding), then compares numeric vectors.

Selected information

Encoding rule

Algorithm now selected

Compare two characters

Character vectors

Each categorical feature contributes equal total weight, even if it has many possible values. This prevents origin or colour from dominating merely because it has more categories.

Characters in the dataset

Untick a character to remove it from the visual map, vector calculations, distances, and clustering. The original CSV is not changed.