Center for Econophysics Research,East China University of Science and Technology A new clustering method for financial time series is proposed,based on the probability densities of the financial assets log return.First,a nonparametric Bayesian method is employed to estimate the probability densities.A measure of discrepancy is then defined between these estimates and the resulting dissimilarity matrix is used to carry out the required cluster analysis.Applying this approach to analyze the N=30 shares composing the Dow Jones industrial average (DJIA) index finds companies that share a similar return behavior.Comparisons are made with other algorithms.