Recent years have witnessed a growing number of researches on community characterization. In contrast to the large body of researches on the categorical measures (rise or decline) for evaluating the community development, we propose to estimate the community development strength (to which degree the rise or decline is). More specifically, given already known categorical information of community development, we are attempting to quantify the community development strength, which is of great interest. Motivated by the increasing availability of large-scale data on the network between entities among communities, we investigate how to score the the community's development strength. We formally define our task as prospecting community development strength from categorization based on multi-relational network information and identify two challenges as follows: (1) limited guidance for integrating entity multi-relational network in quantifying the community development strength; (2) the existence of selection effect that the community development strength has on network formation. Aiming at these challenges, we start by a hybrid of discriminative and generative approaches on multi-relational network-based community development strength quantification. Then a network generation process is exploited to debias the selection process. In the end, we empirically evaluate the proposed model by applying it to quantify enterprise business development strength. Experimental results demonstrate the effectiveness of the proposed method.

翻译：近年来，关于社区特征描述的研究日益增多。与大量采用分类度量（如上升或衰退）评估社区发展的研究不同，我们提出评估社区发展强度（即上升或衰退的程度）。具体而言，在已知社区发展分类信息的基础上，我们试图量化社区发展强度，这一问题具有重要意义。受社区实体间网络大规模数据日益可及的启发，我们研究了如何对社区发展强度进行评分。我们正式将任务定义为：基于多关系网络信息，从分类中推断社区发展强度，并识别出以下两个挑战：(1) 缺乏关于整合实体多关系网络以量化社区发展强度的指导；(2) 社区发展强度对网络形成存在选择效应。针对这些挑战，我们首先采用判别式与生成式混合方法，基于多关系网络量化社区发展强度；随后利用网络生成过程消除选择过程的偏差；最后，通过将其应用于量化企业业务发展强度，对所提出模型进行实证评估。实验结果表明了该方法的有效性。