Research On Knowledge Graph Modeling Method Based On Railway Fault Text Data

The safe operation of railways has always been one of the most important issues for China National Railway Group in the process of railway operation.Centering on the theme of safety,the railway organization continues to build and improve safety management rules and regulations,safety management system and safety operation monitoring system.Therefore,in actual operation,the railway system continues to produce a large number of fault text data containing important information.The analysis of such information can effectively help railway operating enterprises to understand the occurrence and cause of faults and effectively avoid such problems in the actual operation in the future,and ensure the safe operation of railway.Because this kind of information is mostly unstructured data with text as the main carrier,there are many difficulties in data management application.On the one hand,manual analysis requires a lot of human resources and time,which cannot solve problems timely and efficiently.On the other hand,it is difficult for computers to process unstructured text data.In view of the above information processing difficulties,this paper adopts the relevant technologies of the knowledge graph,and uses models such as machine learning and deep learning to realize intelligent mining and analysis of railway fault text data,and evaluate and test through the fault text data generated in actual railway production,building an intuitive and reliable railway fault knowledge system,which provides new ideas and methods for the management and use of railway fault text.The main research contents of this paper are as follows:(1)Based on the named entity recognition of railway fault text data,a BERTBi LSTM-CRF named entity recognition model of railway fault text data is constructed.In this paper,by improving Bil STM-CRF model,BERT-Bil STM-CRF model is constructed to extract valuable train number information,fault component,fault cause,fault nature and fault system information from railway fault text data.The BERT Chinese pre-trained railway fault text data word vector is used as the input of the Bi LSTM-CRF model,so as to automatically capture the feature information from the railway fault text and complete the railway fault named entity recognition task.The comparative experiments on the dataset show that the BERT-Bi LSTM-CRF model achieves more than87% of the named entity recognition indicators in the railway fault text data,and achieves good results.(2)Aiming at the classification of railway fault text data,a BERT-SVM railway fault text data classification model is constructed.The general text classification process requires text data preprocessing,text feature extraction,feature vector representation and vector stitching.The model constructed in this paper realizes the extraction of the main features of the text and the vector representation of the text through the BERT layer,which effectively avoids the error caused by manual extraction of text features and simplifies the related process.At the same time,the model can automatically classify the railway fault text data according to the fault cause,and can also analyze the cause of the fault while saving labor costs.In the process of finding the optimal hyperparameters suitable for the railway fault text classification model,several sets of different parameters are used for comparative experiments.The experimental results show that the railway fault text classification model can achieve better results when C=5 and gamma=0.05.Finally,compared with other classification models,it is proved that the proposed text classification model has better classification performance.(3)Entity relation extraction and knowledge graph construction based on railway fault text data.Aiming at the structural characteristics of railway fault text data,a method based on matching and rules is proposed to extract the relationship of railway fault knowledge entities.The railway fault knowledge entity database is constructed,the relationship between the railway fault knowledge entities is defined,and the acquisition of the railway fault knowledge entity relationship is completed by the method of matching and rules.The collected data is brought into the method to verify,and the results show that the method can effectively obtain knowledge triples from the railway fault text data,and it has high feasibility.Finally,the acquired railway fault knowledge triples are sorted and stored in the Neo4 j graph database,and the Neo4 j visualization platform is used to realize the construction and visual display of the railway fault knowledge graph,so as to help railway operators better grasp the key fault information.Provide accurate and effective information data and technical support for decision-making.