Cluster a Correlation Matrix (in python) – Wil Yegelwel

Abstraction: Reordering correlation matrix columns/rows using hierarchical clustering

Key points:

• cluster_corr() function rearranges a pandas DataFrame or numpy ndarray correlation matrix so highly correlated variables appear adjacent
• Uses scipy.cluster.hierarchy: computes pairwise distances via distance.pdist, builds linkage with method='complete', cuts at pairwise_distances.max()/2
• np.argsort(idx_to_cluster_array) produces the sorted index used to reorder both rows and columns
• Works as drop-in preprocessing for seaborn.heatmap(cluster_corr(df.corr())) to produce interpretable correlation heatmaps

Connections: Scipy · Hierarchical Clustering · Correlation Analysis

Source: https://wil.yegelwel.com/cluster-correlation-matrix/