Clustering stock market companies

The problem of clustering consists in grouping a set of objects in classes, according to their degree of "similarity". A number of criteria can be used to define the intuitive concept of similarity, leading in general to different partitions. Due to this arbitrariness, clustering is an inherently ill-posed problem, as a given data set can be partitioned in many different ways without any particular reason to prefer one solution to another. It is clear that the particular clustering technique can be profoundly influenced by the strategy adopted by the observer and his/her own ideas and preconceptions on the problem.We give a broad introduction to this topic and then describe a few clustering ansatz. We finally apply our results to the analysis of financial time series (Dow Jones Stock Market).