This project leverages machine learning techniques to detect exoplanets from astronomical datasets. By analyzing stellar light curves for periodic dips, the model identifies potential planetary transits with high accuracy, aiding in the discovery of planets beyond our solar system.
With advancements in space exploration and the abundance of astronomical data from missions like Kepler and TESS, the need for efficient and automated methods for exoplanet detection has grown. This project aimed to apply machine learning to streamline and enhance the detection process, reducing reliance on manual analysis.
Light Curve Analysis: The model processes light curve data to detect periodic brightness dips caused by planetary transits.
Binary Classification: A supervised machine learning model predicts whether a given light curve represents an exoplanet transit or noise.
Noise Filtering: Preprocessing steps effectively remove stellar variability and instrumental noise to improve prediction accuracy
Machine Learning Algorithms: Gradient Boosting (XGBoost) and Random Forest for classification tasks.
Preprocessing Tools: FFT (Fast Fourier Transform) and wavelet analysis for signal denoising and feature extraction.
Languages & Libraries: Python, TensorFlow, NumPy, pandas, and Matplotlib for data processing, model building, and visualization.
Dataset: NASA's Kepler Mission light curve data.
Impact
The project achieved a significant improvement in exoplanet detection accuracy compared to baseline methods. It demonstrated the potential of machine learning in transforming astronomical data analysis, providing a scalable approach to exploring the universe.
Takeaways
This project deepened my understanding of signal processing, supervised learning, and data preprocessing techniques. It also strengthened my ability to work with large datasets, reinforcing my passion for applying technology to solve complex scientific problems.