恒星天文光谱数据分类方法探究

Research on Classification Method of Stellar Astronomical Spectral Data

LAMOST-DR7恒星天文光谱数据的用途广泛,其中光谱型数据共有7个类型。笔者对天文光谱数据采取特征选择或降维的方式,探究朴素贝叶斯、逻辑回归、C4.5决策树算法在进行多少降维和特征选择后对该光谱数据的分类效果最佳。对比结果显示,当选择特征数目为101时,softmax模型对天文数据的分类效果最佳。

LAMOST-DR7 stellar astronomical spectrum data has a wide range of uses,of which there are 7 types of spectral data.The author adopts feature selection or dimensionality reduction methods for astronomical spectral data,and explores how many dimensionality reduction and feature selection algorithms of naive Bayes,logistic regression,and C4.5 decision tree algorithms have the best classification effect on the spectral data.The comparison results show that when the number of selected features is 101,the softmax model has the best classification effect on astronomical data.