ISRO Exoplanet ML

Project Architecture & Overview

Engineered for ISRO BAH 2026 Problem Statement 07. Processes NASA Kepler and TESS photometric light curves, performs Box Least Squares (BLS) periodogram search, detrends stellar noise, and classifies true planetary transit dips versus astrophysical false positives using a deep 1D-Convolutional Neural Network.

Key Engineering Milestones

• ✦1D-CNN architecture achieving 94.8% classification accuracy on Kepler confirmed targets
• ✦Automated Box Least Squares (BLS) transit period discovery pipeline
• ✦Robust stellar flare detrending and limb-darkening profile fitting
• ✦Full validation suite with phase-folded transit curves and ROC analysis