一种基于改进分层置信规则库的社交账户可信度评估方法

社交账户可信度评估是确保网络社交生态良性发展的重要环节。针对社交账户可信度评估指标多维、数据信息不确定性多样等问题，提出了一种基于改进分层置信规则库的可信度评估方法。首先从账户属性、交际属性和内容属性三个角度分析了可信度评估各指标之间的相互关系，并依此构建了置信规则库的分层结构。其次，在信息转换函数中引入了自适应系数以更好描述和处理指标间的特性差异。最后，为了弥补专家知识局限性带来的模型误差，采用带有投影算子的协方差矩阵自适应进化策略对自适应系数和模型参数进行了优化。以新浪微博账户作为实验对象，结果表明该方法能够在数据样本有限的情况下获得更高的可信度评估精度。

Credibility evaluation method for social accounts via improved hierarchical belief rule base

Social account credibility evaluation is an important link to ensure the benign development of network social ecology. Aiming at the problems of multi-dimensional credibility evaluation indexes and various data information uncertainty, this paper proposed a credibility evaluation method based on the improved hierarchical belief rule base. Firstly, this paper constructed a hierarchical structure by analyzing the relationship between the indicators of credibility evaluation from three perspectives, Such as account attribute, communication attribute, and content attribute. Secondly, this paper introduced an adaptive coefficient into the information transformation function to better deal with the characteristic differences between indicators. Finally, to make up for the model error caused by the limitation of expert knowledge, this paper used the covariance matrix adaption evolution strategy with projection to optimize the adaptive coefficients and model parameters. Taking Sina Weibo account as the experimental object, the results show that this method can obtain higher accuracy when the data samples are limited.