Predicting Galaxy Redshifts Using Machine Learning Techniques

Accurate estimation of galaxy redshifts is crucial for understanding the universe's structure and evolution. Traditionally, redshifts are measured using spectroscopic methods. These methods are precise but become inefficient and costly when dealing with large datasets generated by modern astronomical surveys. Photometric redshift estimation offers a more efficient alternative. However, traditional photometric methods often depend on spectral energy distribution (SED) templates, which may not be suitable for large volumes of data. To address these challenges, we applied machine learning techniques for photometric redshift estimation. Our study used the GalaxiesML dataset, composed of five photometric magnitudes (grizy) and corresponding spectroscopic labels (green, red, near-infrared, far-infrared, infrared at a longer wavelength than the z band). We trained three machine learning models: K-Nearest Neighbors (KNN), Random Forest (RF), and Multi-Layer Perceptron (MLP). The results were promising. KNN achieved a root mean square error (RMSE) of 0.27, RF achieved an RMSE of 0.26, and MLP had the lowest RMSE of 0.22. These findings highlight the potential of machine learning to efficiently manage astronomical datasets. It offers a scalable and cost-effective method for photometric redshift estimation. This advancement not only helps handle the data deluge in modern astronomy but also enhances our ability to explore the universe's distant regions. Overall, it contributes significantly to the field of astrophysics.