Résumé : We introduce a new unsupervised learning problem : clustering wide-sense stationary ergodic processes and locally asymptotically self-similar processes. Covariance-based dissimilarity measures and asymptotically consistent algorithms are designed for clustering in offline and online data settings, respectively. We discuss an approach to improve the efficiency of clustering algorithms when they are applied to cluster self-similar processes. In a simulation study, several excellent examples are provided to show the efficiency and consistency of the clustering algorithms. Comparison to the other existing clustering algorithms is made. In a real world project, we successfully apply these algorithms to cluster the global equity markets of different regions.