Case Linkage Analysis Based On Logical Regression

With the rapid development of social economy,the transportation and communication,the flow of people,money and object have been increased hugely,the flee-about crime and continuous crime have become more and more prominent,and the crime is more and more professional and specialized.In such cases,the usual practice of organizing an investigation after a case cannot meet the need of actual struggle,so the incorporated investigation came into being.It is a kind of investigation in which the detective organs and investigators incorporate a series of cases which happened continuously in the same or different regions and are proved to be committed by the same criminal or criminals and investigate them together.The premise of the incorporated investigation is "case linkage",that is,to make sure that the suspects in those cases are the same.Case linkage is crucial to the incorporated investigation,and it has a great impact the effectiveness and efficiency of the investigation.In many cases,the specific information(such as fingerprints,fibers,DNA,etc.) does not exist,only some action information of the scene of the crime(such as modus operandi,the time and place of the crime,etc.) The cornmonly used method for case linkage by the legal department is to check the scene annually to confirm the records information,or to look for similar cases on the intemet and compare them,all these methods are very time-consuming.So the goal of case-linkage is to find an automatic and accurate method for incorporated investigation,so that time spent by law enforcement officials on case-linkage can be reduced.Researchers at home and abroad have put forward a number of methods for auto-correlation, but they have not been widely used and they need to calculate the specific weights of all kinds of attributes.However,which is constantly changing in many cases.These changes cause great difficulty for case-linkage.Based on the above analysis,this paper presents a method of auto-linkage based on the logical regression,and this method requires no calculation of the weights of any attributes.This paper first choose attributes useful for case-linkage according to the FBI's NIBRS and suggestions given by experts on crime analysis,then calculates the degree of similarity between different attributes,and then by using the logical regression,determine which attributes can be used to distinguish whether two cases are related.Finally,a test is made on a burglar dataset by using ROC analysis.And the result shows that this method is highly accurate.