Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/LauraSilRu/exoplanet-profiler/llms.txt

Use this file to discover all available pages before exploring further.

ExoProfiler is an unsupervised machine learning pipeline that mines the NASA Exoplanet Archive to uncover hidden structure among distant worlds. Starting from 910 TESS-confirmed planets and 12 physical and orbital features, the pipeline produces two stable, interpretable planetary families — without any predefined labels or categories.

Quickstart

Set up the environment and run the full pipeline end-to-end in minutes.

Pipeline Overview

Understand the six-stage ML workflow from raw data to cluster profiles.

Preprocessing Notebook

Learn how missing values, outliers, and scaling are handled in notebook 03.

Planetary Families

Explore the two discovered families: Standard Population and Exotic Objects.

How It Works

ExoProfiler follows a reproducible six-stage pipeline that takes raw astronomical measurements and transforms them into scientifically interpretable planetary clusters.
1

Acquire & Understand Data

Download the TESS subset from the NASA Exoplanet Archive and run notebooks 01_data_understanding.ipynb and 02_eda.ipynb to explore the 910-planet dataset with 355 columns.
2

Preprocess

Run 03_preprocessing.ipynb to select 12 candidate features, filter planets with fewer than 8 observed values, apply log1p transforms, RobustScaler, and KNNImputer. Output: 731 clean planets.
3

Reduce Dimensions with PCA

Run 04_pca_and_cluster_profiling.ipynb to compress 12 features into 4 principal components that retain 90.35% of total variance.
4

Cluster & Evaluate

Run 05_clustering_and_evaluation.ipynb to train K-Means with K=2 (Silhouette score: 0.4950), validate stability across multiple seeds, and cross-check with DBSCAN and hierarchical clustering.

Key Results

731 Exoplanets

Retained after completeness filtering from 910 TESS-observed planets.

90.35% Variance

Captured in 4 PCA components from 12 original features.

2 Planetary Families

Standard Population (91.24%) and Exotic Objects (8.76%), stable across seeds.

Explore the Documentation

Dataset & Source

NASA Exoplanet Archive TESS subset — feature descriptions and data provenance.

Feature Variables

The 12 planetary and stellar variables used as model inputs.

PCA Component Loadings

How each variable contributes to the four principal components.

Evaluation Metrics

Elbow method, Silhouette scores, and multi-seed stability analysis.

Artifacts & Outputs

All exported CSV, joblib, and JSON files produced by the pipeline.

Limitations

Known data biases, imputation risks, and interpretation caveats.

Build docs developers (and LLMs) love