In this work, we present a machine learning (ML) clustering algorithm for the classification of the interstellar medium (ISM) main components (HII regions, supernova remnants (SNR) and diffuse ionize gas (DIG) regions). We study the ISM components of the NGC 300 galaxy from MUSE integral field spectroscopy observations. These observations give us an ISM spatial resolution of a few parsecs. In order to disentangle this complex ISM, we apply an unsupervised Bayesian Gaussian Mixture Model algorithm to a data set of spaxel-by-spaxel main strong emission lines. Our method produces an automatic and unbiased detection of the main components of the ISM combining the spatial and spectral information.