An Improved Weighted Bayesian Networkand Its Application

Bayesian network(Bayesian Network,BN)is a graphical way to represent the probability uncertainty between variables,and it is an important model to deal with the uncertainty problem.Bayesian networks are widely used in financial analysis,medical diagnosis,machine learning and other fields,and have achieved great success.As a constrained Bayesian network,the Naive Bayesian network has the characteristics of simple structure and high operating efficiency.It has a wide range of applications in the fields of risk analysis and classification.Due to its strong conditional independence assumption in reality It is difficult to achieve in the world.Therefore,how to improve the assumption of conditional independence,learn Bayesian network parameters quickly and accurately from data,and apply the model to practice has attracted more and more attention from scholars.Based on the research of domestic and foreign algorithms,this paper has done the following work on the related research of naive Bayesian network and its application in the evaluation of terrorist attacks:(1)Aiming at the assumption of conditional independence of naive Bayes,different weights are assigned according to the influence of different attributes on classification,so that naive Bayes is extended to weighted Bayes,and the information flow that measures causality is used as the calculation criterion of weight,combined with packaging Method and filtering method,based on genetic algorithm,proposed an improved weighted Bayesian network model(Weighted Bayesian network based on information flow and genetic algorithm,IFG-WBN).The simulation experiment of life expectancy assessment of lung cancer patients after surgery is also used to verify that the model has high validity and accuracy.(2)Aiming at the characteristics of multiple sources,complex relationships and uncertain information in terrorist attacks,as well as the effectiveness and practicality of weighted Bayesian networks based on information flow and genetic algorithm improvements.First,the global terrorism database is used as the original data set,the data is preprocessed,and the random forest is used to filter the indicators to determine the weighted Bayesian network nodes.Secondly,the structure determination and parameter learning are carried out,and then the index weights are calculated based on the information flow,and then the genetic algorithm is used.Determine the optimal weight,evaluate it with probability,and finally test the effectiveness of the IFG-WBN model in practical applications through comparative experiments,so as to better support decision-makers in combating terrorism,thereby reducing the impact of terrorist attacks.The experimental results show that the weighted Bayesian network based on information flow and genetic algorithm is more suitable for dealing with evaluation indicators of complex interrelationships,mining and quantitatively and intuitively expressing the causal relationship between indicators,reasoning based on probability distribution,and the evaluation process is more rigorous,and the accuracy of the assessment results has reached 98.652%,which can more comprehensively reflect the classification of terrorist attack risks.