See this Stack overflow answer for sitting a multivariate Gaussian HMM: https://stackoverflow.com/questions/52141332/is-it-possible-to-fit-a-multivariate-gmhmm-in-hmmlearn. But I wouldn’t recommend using all 280 features. For even 1 state, that would require estimating a 280×280 covariance matrix. Since the covariance matrix is symmetric, that requires the estimation of 280*279/2 = 39,060 covariances and you only have 1106 observations! You should choose only the most informative features.

]]>