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 uses real observational data from the NASA Exoplanet Archive, filtered to planets discovered by the TESS (Transiting Exoplanet Survey Satellite) mission. The dataset captures diverse planetary and stellar characteristics measured across multiple space missions, and serves as the single input to the entire pipeline.
Source
The dataset is drawn from the Planetary Systems (PS) table of the NASA Exoplanet Archive, filtered to TESS discoveries with a single representative row per planet.
| Property | Value |
|---|
| Provider | NASA Exoplanet Archive — Planetary Systems (PS) table |
| URL | https://exoplanetarchive.ipac.caltech.edu |
| Row filter | default_flag = 1 AND disc_facility LIKE '%TESS%' |
| Local path | data/raw/exoplanets.csv |
Raw Dataset Dimensions
The raw file as used in the validated pipeline run contains the following:
| Property | Value |
|---|
| Rows (exoplanets) | 910 |
| Columns (total) | 355 |
| Identifier column | pl_name — no nulls, no duplicates |
| Model feature columns | 12 (selected during EDA) |
pl_name is used exclusively as a row identifier and is never passed to the model.
Completeness Characteristics
Not all 355 columns carry scientific signal relevant to clustering. The table below shows the null rates for each of the 12 selected feature variables across the 910-row raw dataset.
| Variable | Missing (%) |
|---|
pl_insol | 57.58% |
pl_bmasse | 31.76% |
pl_orbeccen | 31.32% |
pl_eqt | 26.26% |
st_met | 24.51% |
pl_orbsmax | 22.53% |
st_mass | 13.30% |
st_logg | 9.56% |
st_teff | 2.53% |
pl_rade | 1.21% |
st_rad | 0.33% |
pl_orbper | 0.11% |
Completeness Filtering
To avoid imputing near-empty profiles, ExoProfiler applies a row-level completeness filter before any transformation: a planet must have at least 8 of 12 features observed to be retained. This caps the imputation burden at four values per row.
| Metric | Value |
|---|
| Original rows | 910 |
| Retained after filter | 731 (80.33%) |
| Excluded rows | 179 |
| Complete cases (no imputation needed) | 249 |
| Cells imputed within retained sample | 9.99% |
preprocessing_row_audit.csv, making the filter fully auditable.
Temporal Bias
Exoplanets discovered in 2026 have a retention rate of only 33.92% after the completeness filter — substantially lower than earlier years. These are likely recent entries whose measurements are still being consolidated in the Archive. This introduces a temporal bias: the 731-planet working sample skews toward planets with more complete, historically stable measurements.
Discovery Method Bias
TESS uses the transit photometry method, which detects planets by measuring the brightness dip when a planet crosses in front of its host star. Transit surveys preferentially detect large planets in close, short-period orbits, because these produce the strongest, most frequent signals. This geometric selection effect shapes the underlying distribution of the dataset and should be kept in mind when interpreting cluster boundaries and the relative sizes of each planetary family.