Live research dashboard
Mission status and cluster quality
Score
0
Quality, speed, and insight
Silhouette
—
Higher separation is better
Davies–Bouldin
—
Lower values are better
Inertia
—
Within-cluster variation
Outliers
0
Generated or detected
Research level
1
0 / 500 XP
Game mode
Choose a planetary research mission
⏱
Guided Orbit Survey
Generate a dataset, select useful features, and create well-separated planetary clusters.
Target: silhouette ≥ 0.45
Hints used: 0
Synthetic observatory
Generate and prepare planetary data
Feature laboratory
Select and weight planetary characteristics
Clustering engine
Configure an algorithm and run the experiment
Fast centroid-based clustering for compact groups.
Algorithm ready.
Run the model immediately, or use step mode to observe assignments and centre movement.
Canvas observatory
Explore planets and cluster assignments
Hover for details. Click a planet to inspect it. In manual mode, drag planets into the coloured wells.
Scientific interpretation
Inspect planets, rename clusters, and explain results
Select a planet on the canvas to inspect its measurements.
Detailed explanations earn interpretation points during mission submission.
Prediction round
Assign an unseen world to a discovered cluster
Generate a mystery planet after clustering.
The result will show your choice, the nearest cluster, distance, and confidence.
Plotly analytics
Visualise clusters, metrics, features, and composition
Evaluation report
Cluster profiles, feedback, and stability
| Cluster | Size | Temperature | Orbit | Radius | Habitability |
|---|
Generate data and run an algorithm to receive contextual guidance.
Progression
Achievements and research milestones
Accessibility and interface
Display, motion, sound, and export options
Learning guide
Understand the clustering concepts used here
Core ideas
- Clustering discovers groups without predefined class labels.
- Classification assigns new records to known categories.
- Scaling prevents large-valued measurements from dominating distance.
- Cluster centres summarise typical members of each group.
- The elbow curve helps compare reasonable cluster counts.
Interpretation cautions
- Different algorithms can produce equally defensible groupings.
- Outliers may be rare discoveries rather than bad data.
- Irrelevant features can hide useful planetary structure.
- Quality metrics support decisions but do not replace scientific judgment.
- Synthetic game data should not be treated as real astronomy.