[Data mining of cataclysmic variables candidates in massive spectra]

Area of Science:

• Astronomy and Astrophysics
• Data Science
• Machine Learning

Context:

• The Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) generates massive amounts of spectral data.
• Efficiently processing this data is crucial for astronomical discovery.
• Identifying specific celestial objects like cataclysmic variables (CVs) requires robust analytical methods.

Purpose:

• To develop an automatic and efficient method for reducing and analyzing LAMOST's spectral data.
• To utilize machine learning techniques, specifically Principal Component Analysis (PCA) and Support Vector Machines (SVM), for spectral data analysis.
• To identify new candidates of cataclysmic variables (CVs) within the LAMOST dataset.

Summary:

• The proposed method uses PCA to construct a feature space from identified cataclysmic variable templates.
• Support Vector Machines (SVM) are employed to exclude non-candidate spectra, followed by a template matching strategy for final candidate selection.
• The identified candidates are used to refine the templates, creating a feedback loop for improved accuracy.

Impact:

• Successfully identified 58 new cataclysmic variable candidates.
• Demonstrates the practical applicability of the developed approach for discovering special celestial bodies in large astronomical datasets.
• Provides a scalable and efficient tool for spectral data reduction and analysis in astronomical surveys.