基于LAMOST数据与机器学习方法测定大样本恒星年龄

Large Sample of Stellar Age DeterminationBased on LAMOST Data and Machine Learning

为了了解银河系以及追溯其形成历史,需要对分布在银河系中的大量恒星样本进行准确的年龄预测。通过LAMOST DR5和Kepler的星震学数据交叉匹配获得的训练样本,给出了一个具有163105颗恒星年龄参数的红团簇星星表。使用核主成分分析与随机森林相结合的方法对多个恒星参数与恒星年龄之间的关系进行训练,将样本分为训练集与测试集进行模型的训练与对照验证,测试集显示所训练的模型对恒星年龄预测的绝对误差平均值为0.46 Gyr,相对误差平均值为13%。同时,还探究了核主成分分析所使用的主成分个数与模型预测性能的关系,结果发现,当主成分达到4个时,模型的预测性能开始趋于稳定。

It is of significant importance to accurately predict the ages of large stellar samples for understanding the Galaxy and tracing its formation history.A catalog of 163105 red clump giants with stellar age label is provided for the train set obtained by cross-matching the LAMOST DR5 data and asteroseismology data of Kepler.The method,a combination of Kernel Principal Component Analysis(KPCA)and random forest,is adopted to train the relationship between multiple stellar parameters and stellar age.The samples are divided into train set and test set for model training and comparison verification.The test set shows that the mean absolute error of trained model for stellar age prediction is 0.46 Gyr,and the mean relative error is 13%.Meanwhile,the exploration of relationship between the principal components used in KPCA and the prediction performance of the model shows that the prediction performance of the model tends to be stable when the number of principal components has reached 4.