Classifying Gamma-Ray Bursts with Gaussian Mixture Model

Using Gaussian Mixture Model (GMM) and Expectation Maximization Algorithm, we perform an analysis of time duration ($T_{90}$) for \textit{CGRO}/BATSE, \textit{Swift}/BAT and \textit{Fermi}/GBM Gamma-Ray Bursts. The $T_{90}$ distributions of 298 redshift-known \textit{Swift}/BAT GRBs have also been studied in both observer and rest frames. Bayesian Information Criterion has been used to compare between different GMM models. We find that two Gaussian components are better to describe the \textit{CGRO}/BATSE and \textit{Fermi}/GBM GRBs in the observer frame. Also, we caution that two groups are expected for the \textit{Swift}/BAT bursts in the rest frame, which is consistent with some previous results. However, \textit{Swift} GRBs in the observer frame seem to show a trimodal distribution, of which the superficial intermediate class may result from the selection effect of \textit{Swift}/BAT.