Average over folds’ confusion matrix for clusters’ members’ classification. Each row of the matrix represents the percentage of samples in an actual cluster, while each column represents the percentage of samples predicted in a cluster. Thus, the normalization of the matrix is such that the sum over columns is 100. Cells with missing annotation contain zero samples. For example, considering the second to last row (Trumpler 20): on average, 90% of members are correctly classified as such, while 10% as members of NGC 6819. Rows and columns in shades of gray indicate clusters not flagged as high-quality.