Self-classification Algorithm And Risk Assessment Of Industrail Safty Accidents

In recent years,the situation of production safety in our country has been gradually improved,and accidents in production safety have been decreasing year by year.However,the situation remains very grim,causing heavy losses to people's lives and property.With the steady progress of work safety supervision,its work must shift from the focus of accident management to risk management.According to the principle of combining prevention with emergency,normal management and abnormal management,the work must be scientifically,normatively and systematically Dynamic hidden danger prevention and analysis system.The data mining technology is now developed rapidly,largely because of its wide range of applications,starting from the data itself,through data cleaning,data changes,data mining,model assessment and knowledge representation and a series of processes,so that people are right Subjects have a clearer understanding of the direction and control and prevention.At present,there are many solutions to solve the problem of safety production by using machine learning methods,but on the whole,the scope of application is narrow.and the stability of the mode is insufficient and the efficiency is low.In order to solve these problems,this paper proposes an improved GBDT algorithm model,which can produce better results in the field of safety production.This article mainly focuses on the following aspects.1.Studying the current research hotspot and current situation of machine learning method in the field of safety production,comparatively analyzing the current mainstream machine learning algorithm model for solving safety problem.2.According to the traditional Boosting algorithm model,an improved GBDT promotion tree model is proposed and applied to the problem of probabilistic prediction of enterprise hidden troubles,and compared with other algorithms to draw a conclusion.3.From the perspective of text mining,this paper proposes to use the TF-IDF feature algorithm and Word2Vec feature algorithm to classify the security content hidden trouble text content,and compare the two.In this paper,based on the safety monitoring data sets provided by the Safety Supervision Bureau of Ningxia Autonomous Region in recent years,by using the machine learning method and improving the current mainstream Boosting model and text feature model,The survey reports categorize these two specific issues as solutions.Finally,based on the experimental training effect and actual demand,the results of this study are summarized and prospected...