A Hybrid Signal Processing and Machine Learning Pipeline for Exoplanet Transit Detection in Kepler Light Curves

We present an open-source Python pipeline for exoplanet transit detection in Kepler photometric data, combining Savitzky-Golay Filter (SGF) detrending, Box Least Squares (BLS) period search, and Transit Least Squares (TLS) physical modelling with a novel hybrid detrending method termed SGWA (Savitzky-Golay Wavelet Adaptive). The pipeline successfully recovers Kepler-22b with a period of 289.86 days (0.0% error) and radius of 2.25 Earth radii. SGWA achieves 18% higher SNR than SGF (37.1 vs 31.4). A Random Forest classifier trained on 7,000+ NASA KOI catalogue entries correctly classifies all planet measurements with 35.4% separation from false positives. Code available at: https://github.com/shaw430/exoplanet-sgwa-pipeline