基于层次化类别嵌入的电力ICT系统故障分类

A Hierarchical Category Embedding Based Approach for Fault Classification of Power ICT System

为解决电力信息通信客服系统在故障研判时存在故障分类准确率低甚至误分的问题,提出基于层次化类别嵌入的文本分类方法,进行电力信息通信系统故障识别。首先,基于电力信息通信系统故障的用户保修工单文本数据构建电力信息通信系统层次化电力故障标签;其次,提出了基于层次化深层金字塔卷积神经网络和基于层次化中断循环神经网络2种层次化文本分类方法,采用层次化类别嵌入方法逐层进行故障类型分类。实验结果表明,基于层次化深层金字塔卷积神经网络的方法效果最优,可以提供高效、准确的故障识别服务。

To solve the low classification accuracy oreven misclassification issue in fault diagnosis,a text classification method based on hierarchical category embedding is proposed in information and communication technology(ICT)customer service systems.First,a hierarchical label system is constructed for the failure data in power ICT systems based on the textual data of the work orders.Then,hierarchical deep pyramid convolutional neural networks(HDPCNN)and hierarchical disconnected recurrent neural networks are proposed,which adopt hierarchical category embedding technique for level-by-level fault type classification.The experimental results show that the hierarchical text classification algorithm HDPCNN has the best classification accuracy,which can provide efficient and accurate services for fault type recognition.