Hertzsprung and Russell, created a diagram of then known stars with respect to absolute magnitudes or luminosities versus their stellar classifications or effective temperatures. This gave a clear clusters of star types, namely main sequence stars, from birth to maturity, followed by giants, supergiants and white dwarfs. With the rise of technology number of stars with known properties had been growing exponentially and manual categorization is futile. Using the same parameters of HR diagram, this paper analyzes the efficiency of unsupervised ML algorithm expectation-maximization clustering on a database containing 120 000 stars.