Research On Construction And Prediction Of Hidden Danger Network In Logistics Enterprises

The logistics industry has strong mobility,uncertainty and resource aggregation.With the rapid development of the industry,the hidden security problems of logistics enterprises are constantly changing.The problem of safety hazards is not found in time,and once it causes an accident,it will bring huge losses and seriously affect people’s life and property safety.Therefore,this paper starts with the historical inspection data of hidden dangers in logistics enterprises,and focuses on the prevention of hidden dangers in advance to effectively reduce the security risks of logistics enterprises.At present,there have been many researches on hidden danger prediction,but they are mainly concentrated in the fields of transportation,coal mine,electric power,etc.,and most of them are aimed at the analysis of a single hidden danger subject or a single hidden danger class.There are still some deficiencies in the researches on the correlation and structural changes between multiple hidden danger subjects and multiple hidden danger classes.Therefore,based on the graph structure data,this paper uses the association relationship between nodes to build the hidden danger network of logistics enterprises,and realizes the prediction and completion of hidden danger facts of logistics enterprises based on the study of time series knowledge graph representation learning and incremental learning,which mainly includes the following three aspects of research content.1.Construction and analysis of hidden danger network of logistics enterprises.Based on the processed structured hidden trouble data of logistics enterprises,the top-down construction of hidden trouble knowledge graph of logistics enterprises is adopted,and the construction and visualization of hidden trouble network of logistics enterprises are realized from the concept layer design to the instance layer learning.At the same time,based on the node centrality measurement index of complex network analysis,the entity importance analysis of logistics enterprise hidden danger network is completed,so as to promote the hidden danger investigation work more targeted.2.Research on Hidden danger network prediction of logistics enterprises based on time series fusion.The common knowledge graph link prediction ignores the use of time information.Based on the learning technology of hidden danger fact representation of logistics enterprises that integrates time information,this paper uses time sequence encoder and static decoding function to complete the prediction and completion of hidden danger facts,so as to guide the safety management of logistics enterprises with a more comprehensive hidden danger knowledge network.3.Research on Hidden Danger network prediction of logistics enterprises based on incremental learning.Although the full model learning method can obtain more data information,it cannot complete the continuous learning of new knowledge.Hidden danger facts are constantly generated and developed.Considering the excessive occupation of training resources caused by the increase of data volume,this paper comprehensively adopts the incremental learning technology based on playback and regularization on the basis of time series fusion,and realizes the continuous learning and adjustment of the model through the incremental learning mode of hidden danger facts,so as to improve the prediction efficiency and the ability of learning new knowledge of the model.