Data mining techniques on astronomical spectra data. I : Clustering Analysis

TL;DR

This paper reviews clustering techniques for astronomical spectra, compares their performance on standardized datasets, and provides open-source tools to facilitate further research in spectral analysis.

Contribution

It offers a comprehensive review, experimental comparison, and practical resources for clustering methods applied to astronomical spectral data.

Findings

Performance varies significantly across clustering methods.

Spectra from LAMOST and SDSS are used for benchmarking.

Open-source code is provided for reproducibility and further development.

Abstract

Clustering is an effective tool for astronomical spectral analysis, to mine clustering patterns among data. With the implementation of large sky surveys, many clustering methods have been applied to tackle spectroscopic and photometric data effectively and automatically. Meanwhile, the performance of clustering methods under different data characteristics varies greatly. With the aim of summarizing astronomical spectral clustering algorithms and laying the foundation for further research, this work gives a review of clustering methods applied to astronomical spectra data in three parts. First, many clustering methods for astronomical spectra are investigated and analysed theoretically, looking at algorithmic ideas, applications, and features. Secondly, experiments are carried out on unified datasets constructed using three criteria (spectra data type, spectra quality, and data volume) to compare the performance of typical algorithms; spectra data are selected from the Large Sky Area Multi-Object Fibre Spectroscopic Telescope (LAMOST) survey and Sloan Digital Sky Survey (SDSS). Finally, source codes of the comparison clustering algorithms and manuals for usage and improvement are provided on GitHub.