BUKU TUGAS AKHIR: IDENTIFIKASI EKSOPLANET PADA KURVA CAHAYA KEPLER MENGGUNAKAN CONVOLUTIONAL NEURAL NETWORK (CNN)

• Main Author: Naufal, Muhammad
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/69093
• Institution: Institut Teknologi Bandung

The Kepler Space Telescope was launched in 2009 to detect Earth-sized planets using the photometric transit method, surveying the region of the Milky Way Galaxy. Kepler has observed about 200,000 stars and has found more than 2,000 confirmed exoplanets. The mission was also aimed at characterizing planets the size of Earth or smaller in/near the habitable zone. However, these planets are at the tip of the mission’s detection sensitivity, making them a little difficult to identify. Exoplanetary population studies require automation and accuracy in assessing the likelihood that a candidate planet is indeed a planet, even at low signal-to-noise. Astronet is a deep learning model for identifying exoplanets on the star’s light curve. Astronet uses the convolutional neural network model which is already widely used in various subjects. This final project re-examines the model to get results that are in accordance with the initial research. This model can predict whether the signal on the light curve is a planet or a false positive coming from an instrument error or other phenomena. With an accuracy of 0.96 and an AUC score of 0.98, the model is very effective in ranking individual candidates based on the probability that they really are a planet. The model is applied to the identified candidate signal set from the known Kepler planetary system.