Research on airport passenger risk assessment by fusing Bayesian Optimized Random Forest
To scientifically and effectively evaluate passenger risk and enhance the convenience of passenger travel,this study focused on airport departing passengers.Relevant information for assessing passenger risk was gathered from Passenger Name Records and was combined with passenger security information to develop a survey questionnaire.The rationality of valid questionnaire data was tested using SPSS 22.0 software,leading to the creation of a passenger risk assessment index system.Random Forest,an ensemble learning algorithm,generates decision trees from randomly selected sample sets and feature sets.This study used a probabilistic surrogate model to assess hyperparameters within the confidence interval and constructed a Bayesian Optimized Random Forest(BO-RF)to evaluate the risk level of passengers.The decision-making process of the model was explained and analyzed using Shapley additive explanations(SHAP)method.The results show that the key indicators influencing passenger risk assessment are contraband records of checked baggage,the number of annual flights,the number of passengers per Passenger Name Record(PNR),travel date,date of birth,cabin class,quantity of checked baggage,and gender.For assessing low-risk passengers,the number of annual flights is found to be the most critical,followed by contraband records of checked baggage and the number of passengers per PNR.For assessing medium and high-risk passengers,contraband records of checked baggage are the most important,followed by the number of annual flights and the number of passengers per PNR.There is a negative correlation between the number of annual flights and low-risk predictions and a positive correlation between the number of annual flights and medium-risk predictions.Contraband records of checked baggage are positively correlated with low and medium-risk predictions.Compared with many other traditional algorithms,the performance of the Bayesian-optimized random forest is better.The accuracy rate reaches 97％and the consistency K-value is 0.967 7.These findings provide valuable insights for implementing differentiated security checks at airports.
safety social engineeringpassenger riskBayesian Optimized Random Forest(BO-RF)risk level
万方数据
维普
