Clustering financial time series

We analyze the shares aggregated into the Dow Jones Industrial Average (DJIA) index in order to recognize groups of stocks sharing synchronous time evolutions. To this purpose, a pairwise version of the Chaotic Map Clustering algorithm is applied: a map is associated to each share and the correlation coefficients of the daily price series provide the coupling strengths among maps. A natural partition of the data arises by simulating a chaotic map dynamics. The detection of clusters of similar stocks can be exploited in portfolio optimization.