This document discusses clustering financial time series data using correlation matrices. It summarizes that analyzing 560 credit default swaps over 2500 days, the empirical correlation matrix eigenvalues closely match the theoretical Marchenko-Pastur distribution, indicating noise. Only 26 eigenvalues exceed the theoretical maximum, which may correspond to market and industry factors. Hierarchical clustering can reorder assets to reveal correlation patterns. Filtering by this reveals the underlying network structure. Beyond correlations, copulas represent the dependence structure, and a distance measure is proposed combining L1 and L0 distances of cumulative distribution functions to cluster on full distributions rather than just correlations. Stability tests show the proposed approach yields more robust clusters than standard correlation-based methods.