Fig. 2From: Unsupervised software defect prediction using median absolute deviation threshold based spectral classifier on signed Laplacian matrixPredicted clusters result. a, b illustrate an example of the cluster memberships of CM1 dataset, which are constructed using the zero threshold based spectral clustering and MAD threshold based spectral clustering, respectively. Related to Table 4, the use of MAD as partitioning threshold can increase the number of predicted cluster membershipsBack to article page