Research On Security Early Warning And Decision-making Of SL Logistics Park

The logistics park is an important node for building a modern logistics system and an important support for the sustainable development of China's logistics industry.With the rapid increase in the number of logistics parks in recent years,logistics park security incidents have attracted attention.Safety accidents are often harmful and heavy,and safety should be the basic condition for the primary guarantee of production and operation of enterprises.Therefore,the safety management of logistics parks is a key issue to be solved in the operation of the park.Then how to strengthen safety management,safety warning,and eliminate accidents are particularly important for the safe operation of logistics parks.This paper analyzes the characteristics of logistics park security and expounds the problems that are common in the security management of logistics parks in China.Taking the Shaliang Logistics Park in Hohhot,Inner Mongolia(hereinafter referred to as SL Logistics Park)as the research object,this paper introduces the security management mechanism of SL Logistics Park and advises on the main security issues that exist.Based on the theory of accident cause,the direct causes and indirect causes are analyzed to explain the causes of safety accidents in SL logistics parks.According to the construction principle of the evaluation index system,combined with the particularity of the logistics park,a safety early warning evaluation index system was established.Relying on this index system,the genetic algorithm improved neural network(GA-BP)model and the random forest(RF)model are established as the safety evaluation and early warning model of SL logistics park.The models were trained,tested and compared use the objective real data of the SL logistics park,and finally achieved good evaluation and early warning.Considering the reliability and generalization requirements of security early warning,this paper proposes a GA-BP-RF fusion model based on Staking.Compared with a single model,the accuracy and reliability of the fusion model are higher.Based on the GA-BP-RF safety evaluation and early warning model,a sensitive index selection model and an indicator adjustment strategy model are established based on the gradient cosine similarity and Adam(Adaptive Moment Estimation).The sensitive index selection model selects the index that has the greatest impact on the current logistics park security status by comparing the cosine similarity between the index direction and the gradient direction to prompt the security management personnel to focus on monitoring and prioritizing the sensitive indicators.When the system is in an unsafe state,the indicator adjustment strategy model uses GA-BP-RF as the objective function,and uses the Adam algorithm to quickly search for the indicator status corresponding to the adjacent security point,providing a reference for decision-making for security management personnel.Finally,the paper proves the feasibility of the model from security early warning to decision making through case analysis,the GUI(Graphical User Interface)is used to realize the visualization of the early warning and decision model,the calculation results of the model are more intuitively displayed to the security management personnel,improving the practicability and operability of the model.This paper has a total of 21 pictures,17 tables,72 references.