ExoplanetHunter-TransitML

Detect exoplanet transits using classical machine learning on Kepler lightcurve data. The project leverages Lightkurve for lightcurve retrieval and preprocessing, and grouped_timeserie_cv for grouped time series cross-validation.

Installation

pip install lightkurve
pip install grouped_timeserie_cv

Features

• Download and normalize Kepler/TESS lightcurves.
• Segment long lightcurves into fixed-length windows.
• Compute statistical features (mean, std, min, max, skew, kurtosis, transit depth) per segment.
• Label segments as containing a transit (1) or not (0).
• Use grouped cross-validation to avoid data leakage across targets or segments.

Usage Example

import lightkurve as lk
from transit_dataset import build_segmented_dataset

# Download and preprocess a single lightcurve
lc = lk.search_lightcurve("Kepler-10").download().remove_nans().normalize()

# Build segmented dataset
df_segments = build_segmented_dataset(lc, segment_length=200, sigma=5)
df_segments["target_id"] = "Kepler-10"

# Save to CSV
df_segments.to_csv("transit_segments.csv", index=False)

Links

• Lightkurve Documentation
• grouped_timeserie_cv GitHub

---