So far, we have addressed clustering issues through different approaches. Another way of dealing with these problems is the model-based approach. In this case, we will use certain models for clusters and attempt to optimize the fit between the data and the model. In other words, each cluster can be mathematically represented by a parametric distribution, such as Gaussian (continuous) or Poisson (discrete).
Gaussian distribution has some limitations when modeling real-world datasets. Very complex densities can be modeled with a linear combination of Gaussian weights weighed appropriately. A mixture model is a type of density model that is packed with a number of density functions, usually Gaussian (Gaussian Mixture Models (GMM)), and these functions are combined to provide multimodal density. These models allow the representation of probability distributions in the presence of subpopulations, where the mixture components are the...