GMM

Imagine we have a customer database and want to segment those customers based on their purchase history. We’re not sure how many segments there might be, and unlike in classification tasks, there is no real baseline that we can use to compare. This is a typical use case for clustering, where we don’t know the number of clusters in advance and need the algorithm to tell us.

Gaussian mixture model (GMM) is a probabilistic algorithm commonly used for clustering tasks, including customer segmentation. It leverages the concept of a mixture model, which represents the data distribution as a combination of several Gaussian distributions. GMM assumes that the data points are generated from a mixture of Gaussian distributions, with each distribution having its own mean and covariance matrix.