DCNN: Carbon Star Classification Based on Convolutional Neural Network

Carbon star is a rare type of star. In order to solve the problem of slow speed and low accuracy of carbon star classification relying on iterative pattern matching technology, in LAMOST (Large Sky Area multi-object Fiber Spectroscopic Telescope, LAMOST (7th Data Release (DR7)). We selects carbon star spectral data and applies it to deep convolutional neural network. The experiment uses the Sloan Digital Sky Survey (SDSS) and LAMOST spectral data of 2656 carbon stars to train the model. Carbon stars are divided into five subcategories: C-H, C- R, C-J, C-N and Ba. Finally, the classification results are compared with Random Forest (RF), Densenet, Inception V3, and Resnet models using galaxy-based classification accuracy and other indicators. Experimental results show that the accuracy of the model based on convolutional neural network in carbon star classification reaches 82%. It demonstrates that our method is useful for the correct classification of carbon stars.