Modeling And Analysis Of Network Autocorrelation Model About College Students' Bad Behavior Based On Bayes Estimation

College students' bad behavior will bring a great deal of trouble to their own development,school management and others,and also will lead to social instability.Existing research shows that college students' bad behavior is significantly associated with interpersonal relationships.In the field of psychology and pedagogy,research on college students' bad behavior is common,but it lacks further quantitative analysis and discussion.Network autocorrelation model is belong to Social Network Analysis,its benefit is not only considering the impact of personal networks,but also considering their own factors.Network autocorrelation model is a beacutiful network analysis tools.In this paper,influence factors that trigger college students' bad behavior are studied quantitatively by using a network autocorrelation model.The research are reflected herein:Firstly,on the basis of existing research on college students' bad behavior,this paperestablishes the indication system of college students' bad behaviors;Secondly,combined with the Network autocorrelation model's development and college students' bad behavior index system,this paper establishes a multiple networks autocorrelation model about students' bad behavior;Thirdly,respectively expounds the theory based on Ordinary Least Squares(OLS)Estimation? Maximum Likelihood Estimation(MLE)and Bayes Estimation of multiple network autocorrelation model about parameter estimation methods,and improves the inconsistent of OLS estimation.Gives the programming steps about MLE and Bayes Estimation;Finally,empirical analysis is to our College students: includes completes the data collection,analyzes the role of the network through the social network analysis method and network autocorrelation model.Three methods in the empirical comparative discussion.At last gives a reasonable way to help correcting college students' bad behavior.Through empirical analysis in network autocorrelation model about college students' bad behavior,obtain the results are reflected herein:On the one hand,in terms of model parameter estimation,the estimate of OLS estimations is inconsistent because the dependent variable is influenced by a network delay item.In order to get consistent estimators of the parameters,we give a two-stage least squares(2OLS)estimation of the model parameters.The estimate of MLE is unbiased and consistent.The imitation of the MLE has a better result than OLS.But in a large sample size,MLEbehaves badly about the calculation and the computational complexity.In setting the parameter of the prior distribution,it can deduce the conditional distribution of each parameter through Bayes estimation,and it can obtain a convergence and stability of parameter estimation sample by MCMC sampling.This method can reduce the effects of sample data of outliers,and has simple calculation.The results show that the maximum likelihood method,Bayes estimation are higher than the least squares fitting degree;On the other hand,from the empirical results,subjective causes of personal relationships have a significant impact oncollege students' negative emotions and bad behavior.Objective causes of life satisfaction and economic pressure have a significant affect on anxiety,economic pressures and interpersonal stress have a significant effect on depression and bad behavior.It is necessary to consider from both the subjective reasons and objective reasons to correct college students' bad behavior.Research using network autocorrelation models not only have practical value,but also theoretical significance.