Research On Elevator Risk Assessment Based On Bayesian Modified Logistic Regression Model

With the steady growth of economy and the gradual advance of urbanization,the quantity of elevator has been rising.However,it's maintenance and supervision are not good enough so that the elevator accidents still occur frequently.Recently,more and more serious injuries have caused public panic and the safety of elevator operation has become a social focus.After that,many provinces have issued protection and supervision measures.But unclear responsibility management makes it difficult to implement elevator safety responsibilities,and bring regulatory conflicts to the quality inspection.In order to ensure the safe operation of elevators,improve the efficiency of elevator sampling,and put forward new regulatory models,the related spot-check management system and application research are in no hurry.In order to improve the efficiency and effectiveness of elevator inspections,it builds a management system based on the analysis of sampling data,using elevator usage,basic parameters and manufacturing maintenance as indicators.It introduces the existing research,laws and regulations of elevator at home and abroad.According to spot-check data,using Bayesian improved Logistic regression to get new methods-Bayesian Modified Logistic Regression.In the empirical analysis,the data was first preprocessed,and the independent variables were recoded using the Bayesian modified WOE coding method.Then the risk grading was constructed.The Bayesian modified Logistic regression method was used to establish the quantification of the risk of the entire elevator.The model forms an elevator overall risk assessment system.Comparing the two risk models by calculating LIFT statistics and K-S statistics,it is concluded that Bayesian modified Logistic regression is more accurate.Finally,the two models are combined with two sampling methods for simulation.The results show that the sampling ratio of the combination of Bayesian modified Logistic regression and the weighted average risk value method is more reasonable and accurate.In summary,Bayesian correction can improve the prediction accuracy of Logistic regression.