Community detection problem is a well-studied problem in social networks. One major question to this problem is how to evaluate different community detection algorithms. This issue is even more challenging in the problem of local community detection where only local information of communities is available. Normally, two community detection algorithms are compared by evaluating their resulted communities. In this regard, the most widely used technique to evaluate the quality of communities is to compare them with the ground-truth communities. However, for a large number of networks, the ground-truth communities are not known. As a result, it is necessary to have a comprehensive metric to evaluate the quality of communities. In this study, improving a local quality metric, a number of local community detection algorithms are compared through assessing their detected communities. Furthermore, using some small graphs as example communities, some drawbacks of a number of existing local metrics are discussed. Finally, according to the experimental results, it is illustrated that the local community detection algorithms are fairly compared using the proposed metric, GDM. It is also shown that the judgment of GDM is almost the same as that of F1-score, i.e. the metric which compares the community with its ground-truth community.

社区检测问题是社交网络中一个经过充分研究的问题。这个问题的一个主要问题是如何评估不同的社区检测算法。这个问题在只有社区的本地信息可用的本地社区检测问题中更具挑战性。通常，通过评估它们的结果社区来比较两种社区检测算法。在这方面，评估社区质量的最广泛使用的技术是将它们与真实社区进行比较。然而，对于大量网络，真实社区是未知的。因此，有必要有一个全面的指标来评估社区的质量。在这项研究中，改进本地质量指标，通过评估其检测到的社区，比较了许多本地社区检测算法。此外，使用一些小图作为示例社区，讨论了许多现有本地指标的一些缺点。最后，根据实验结果，说明了使用所提出的度量标准 GDM 对本地社区检测算法进行了公平的比较。还表明，GDM 的判断与 F1-score 的判断几乎相同，即比较社区与其真实社区的度量。这表明使用所提出的度量标准 GDM 可以公平地比较本地社区检测算法。还表明，GDM 的判断与 F1-score 的判断几乎相同，即比较社区与其真实社区的度量。这表明使用所提出的度量标准 GDM 可以公平地比较本地社区检测算法。还表明，GDM 的判断与 F1-score 的判断几乎相同，即比较社区与其真实社区的度量。