关键词: 服务价值链, 故障文本分类, 交互注意力机制, 特征融合, 故障诊断

Abstract: Deep learning model has become a key technology for fault diagnosis and analysis.It is difficult to effectively extract key fault text features only by recurrent neural network or convolutional neural network.To solve the problem,a mutual attention mechanism network model was proposed to extract key feature from recurrent neural network and convolutional neural network and then generate global-local features.To further improve classification performance,a faulty parts and faulty mode attention mechanism were proposed to focus on key features in fault text and then generate fault parts and fault mode features.Features used for classification were generated by concatenating these features.The model was used for classifying several fault text data sets.Experiment results showed that the proposed model had a promising result compared with other models.

Key words: service value chain, fault text data classification, mutual attention mechanism, feature fusion, fault diagnosis