From the perspective of psychological contract,this paper discusses mechanism of consumers' network cluster behavior in the context of brand crisis. On the basis of Simmel's conflict theory,it presented new fi...From the perspective of psychological contract,this paper discusses mechanism of consumers' network cluster behavior in the context of brand crisis. On the basis of Simmel's conflict theory,it presented new findings of network cluster behavior. It is concluded that brand crisis exerts significant influence on breach of psychological contract. Particularly,functional brand crisis more easily leads to breach of transactional psychological contract,while value brand crisis more easily leads to breach of relational psychological contract. Breach of transactional psychological contract more easily leads to realistic network cluster behavior,while breach of relational psychological contract does not necessarily lead to non-realistic network cluster behavior.展开更多
电力用户行为数据维度高、特征间存在共线性,导致信息冗余和分析精度下降,影响检测效果。为了有效确保电网的安全和稳定运行,提出一种基于反向传播(Back Propagation,BP)神经网络的电力用户异常用电行为检测方法。构建正常用电行为基准...电力用户行为数据维度高、特征间存在共线性,导致信息冗余和分析精度下降,影响检测效果。为了有效确保电网的安全和稳定运行,提出一种基于反向传播(Back Propagation,BP)神经网络的电力用户异常用电行为检测方法。构建正常用电行为基准模型,刻画时空特征以识别偏离行为。针对特征差异,提出最优特征评价模型,通过优选机制降维、解决共线性问题,用信息熵量化特征贡献度并归一化处理,迭代筛选强判别性特征子集。随后,基于K-means聚类分析用户电力特征数据,实现用户分群。引入随机矩阵理论评估用户行为模式影响因素。用电行为刻画后,采用BP神经网络检测异常,针对高维特征,用子空间聚类算法划分空间,BP神经网络通过迭代优化训练模型,调整权重参数完成检测。实验研究表明,本文方法聚类效果最佳,贝叶斯检出率受迭代次数影响小,稳定性强,检测性能更优;ROC曲线下的面积(Area Under the Curve,AUC)值更接近1。在电力用户异常用电行为检测方面性能良好,可以得到高准确率的检测结果。展开更多
文摘From the perspective of psychological contract,this paper discusses mechanism of consumers' network cluster behavior in the context of brand crisis. On the basis of Simmel's conflict theory,it presented new findings of network cluster behavior. It is concluded that brand crisis exerts significant influence on breach of psychological contract. Particularly,functional brand crisis more easily leads to breach of transactional psychological contract,while value brand crisis more easily leads to breach of relational psychological contract. Breach of transactional psychological contract more easily leads to realistic network cluster behavior,while breach of relational psychological contract does not necessarily lead to non-realistic network cluster behavior.
文摘电力用户行为数据维度高、特征间存在共线性,导致信息冗余和分析精度下降,影响检测效果。为了有效确保电网的安全和稳定运行,提出一种基于反向传播(Back Propagation,BP)神经网络的电力用户异常用电行为检测方法。构建正常用电行为基准模型,刻画时空特征以识别偏离行为。针对特征差异,提出最优特征评价模型,通过优选机制降维、解决共线性问题,用信息熵量化特征贡献度并归一化处理,迭代筛选强判别性特征子集。随后,基于K-means聚类分析用户电力特征数据,实现用户分群。引入随机矩阵理论评估用户行为模式影响因素。用电行为刻画后,采用BP神经网络检测异常,针对高维特征,用子空间聚类算法划分空间,BP神经网络通过迭代优化训练模型,调整权重参数完成检测。实验研究表明,本文方法聚类效果最佳,贝叶斯检出率受迭代次数影响小,稳定性强,检测性能更优;ROC曲线下的面积(Area Under the Curve,AUC)值更接近1。在电力用户异常用电行为检测方面性能良好,可以得到高准确率的检测结果。
