Research On Security Event Association Analysis Method Based On Semantic Entity Recognition

At present,the continuous development of data mining technology plays an important role in social governance and social management.In many fields of society,such as social media platforms,enterprise information systems,a large number of data related to safety activities or safety production have been generated.These data show the characteristics of quantification,multi-source and isomerization.It is of great theoretical significance and practical value to analyze and study these safety production data,identify the relevant factors affecting safety activities and explore the mechanism of unsafe activities.The research project takes the text data of accidents in production safety activities as the research object,takes finding the correlation relationship of accident factors as the main research goal,and aims to provide technical support and theoretical basis for the emergency management and accident analysis of production safety.The research content of this project mainly includes the following aspects:(1)Aiming at the lack of annotated corpus in the security field,a named entity recognition model based on feature template and deep learning is proposed.Firstly,aiming at the entity information with strong features and less quantity in text data,a feature template method is designed for recognition.Secondly,the transfer learning model is constructed to recognize the entity information which has no obvious characteristics but has a large number.Finally,edit distance is introduced to calculate the similarity between entities to complete the entity disambiguation task.Compared with the current mainstream deep learning method,the model proposed in this thesis has higher accuracy in entity recognition task in the field of production safety.(2)In order to analyze the association relationship between accident factors in production safety activities,a method of association rule discovery based on joint topic causality analysis is proposed.Firstly,on the basis of semantic entity recognition,event information is further extracted by extracting event trigger words to match event arguments.Secondly,the topic of the accident text is analyzed,and the optimal topic classification is obtained by synthesizing a variety of classification results.Finally,the causal analysis is used to explore the correlation between various accidents and risk factors.The experimental results show that the proposed method can effectively analyze the accident factors of production activities,and find out the key factors affecting production activities.(3)Based on the accident data of power grid,a prototype system of safety production accident analysis is designed and implemented.The system includes entity recognition,event correlation analysis and visual display of power grid accidents.The system has good domain data independence,and can be easily applied to text analysis and data mining tasks in the field of work safety.There are 33 pictures,36 tables and 88 references.