Application of random forest to stellar spectral classification

Classifying stellar spectra is an important work in astronomy. Numerous automated classification techniques have been explored for spectra data classification. But achieving high accuracy of spectral classification is still a goal of study. Random Forest is a recently available ensemble learning algorithm. Existing literatures have shown the superior performance of random forest in a few application areas. In this paper, random forest is used to approximate stellar spectral classification from stellar spectra. Our objective is to evaluate effectiveness of random forest on classifying stellar spectra. An experiment of performance comparison between random forest and multilayer perceptron network shows that the former one has a better efficiency and less RMS error.