V-measure is an external cluster evaluation measure based on conditional entropy that addresses several issues with existing cluster evaluation methods. It combines homogeneity and completeness using a harmonic mean, similar to the F-measure in information retrieval. Homogeneity measures how well clusters contain only members of a single class, while completeness measures how well all members of a class are assigned to the same cluster. V-measure normalizes these measures and combines them to provide a more comprehensive evaluation of clustering solutions. It is invariant to the number of classes, clusters, and data points, making it applicable across various clustering tasks. V-measure has been compared to other evaluation measures like Purity, Entropy, and F-measure, and it has been shown to better handle the "problem of matching" by evaluating the entire membership of clusters rather than just matched portions. It has been applied to document clustering and pitch accent type clustering, demonstrating its effectiveness in evaluating clustering success. V-measure provides a more accurate and comprehensive assessment of clustering solutions by considering both homogeneity and completeness, making it a valuable tool for evaluating clustering performance.V-measure is an external cluster evaluation measure based on conditional entropy that addresses several issues with existing cluster evaluation methods. It combines homogeneity and completeness using a harmonic mean, similar to the F-measure in information retrieval. Homogeneity measures how well clusters contain only members of a single class, while completeness measures how well all members of a class are assigned to the same cluster. V-measure normalizes these measures and combines them to provide a more comprehensive evaluation of clustering solutions. It is invariant to the number of classes, clusters, and data points, making it applicable across various clustering tasks. V-measure has been compared to other evaluation measures like Purity, Entropy, and F-measure, and it has been shown to better handle the "problem of matching" by evaluating the entire membership of clusters rather than just matched portions. It has been applied to document clustering and pitch accent type clustering, demonstrating its effectiveness in evaluating clustering success. V-measure provides a more accurate and comprehensive assessment of clustering solutions by considering both homogeneity and completeness, making it a valuable tool for evaluating clustering performance.