Statistical Modeling Of Acquaintance Injury In College Students And Association Rules Mining

College students are one special group of people during their fast-growing up period and usually draw attention of whole nation and society.They are young and active,passionate,impulsive,but not mature enough.There is a highly likelihood for them to encounter injury incidents,especially among acquaintances.Studying the problem of acquaintance injury may provide relevant theoretical foundations and suggestions of countermeasure for reducing and preventing injury incidents occurring on campus.This dissertation analyzes descriptively the problem of acquaintance injury by using 6886 data samples from 6 universities in Yunnan province and shows that the acquaintance injuries take place about 58% of the total injury incidents happened on campus.Moreover,the result of the acquaintance injuries is worse than those of self-inflicted and stranger-inflicted injury.By using statistical modeling and data mining methods respectively,the influence factors and the injury characteristics of the acquaintance injury are analyzed.The main obtained results are given as follows:Firstly,based on the described analysis of the data of samples,some basic injury characteristics is obtained.For example,the students who are male or sophomore students,whose families' allowance per-month is less than 1500 RMB,or who are from double-parent families,may be hurt more from acquaintance than others.Secondly,the main influence factors of the acquaintance injury are extracted by applying the methods of Logistic regression,stepwise regression.The ordering degrees of importance of the influence factors are given as follows: grade,school,the relationship with the supervisor,negative emotions,parents' education level,in a romantic relation or not,families' allowance per-month,absenteeism for classes,school awards,exercise,gender,ability of preventing injury,gambling,few security awareness,emotional maturity,nationality,instance of getting along with roommates,taking drugs or not.By combining a total of 6 models of three versions of the random forest(the ntree equals to 200,500 and 700),regression trees in two versions(the SE is equal to 0.5 and 1,respectively),and Logistic models,then conducting cross-validation method,it is found that the optimal version of the statistical model for acquaintance injury is the random forest with ntree=700.Thirdly,by applying the association rules and R programming,the relationships of multivariable have been revealed as well as the characteristics of acquaintance injury incidents related to month,time period,place and population.The characteristics of month shows that the injury incidents take place primarily in the first six months of a year,while July is also a period when such incidents occur frequently in the second half year.Moreover,the characteristics of time period shows that the injury incidents are normally take place in students' leisure time from 17:30 to 00:30;especially during 17:30-19:00 when there is a great possibility,and that the risk reaches the peak during 00:00-00:30.The characteristics of place shows that the acquaintance injury incidents occurs mostly in dormitories and the playing fields on campus and the internship workplace and the way go out and back out of campus.The characteristics of population show that the students from freshmen to doctoral students of each grade have their own injury characteristics.According to these characteristics,a number of recommendations are put forward.Finally,a real example which occurred in a university and had greater social influence is analyzed and verified.