Data Mining to Criminal Career Analysis

The amount of data being produced in modern society is growing at an accelerating pace. New problems and possibilities constantly arise from this so-called data explosion. One of the areas where information plays an important role is that of law enforcement. Obviously, the amount of criminal data gives rise to many problems in areas like data storage, data warehousing, data analysis and privacy. Already, numerous technological efforts are underway to gain insights into this information and to extract knowledge from it.

This paper discusses a new tool that attempts to gain insights into the concept of criminal careers: the criminal activities that a single individual exhibits throughout his or her life. The national police annually extracts information from digital narrative reports stored throughout the individual departments and compiles this data into a large and reasonably clean database that contains all criminal records from the last decade. From this database, our tool extracts the four important factors (see Section II) in criminal careers and establishes a clear picture on the different existing types of criminal careers by automatic clustering. All four factors need to be taken into account and their relative relations need to be established in order to reach a reliable descriptive image. To this end we propose a way of representing the criminal pro_le of an individual in a single year. We then employ a speci_cally designed distance measure to combine this pro_le with the number of crimes committed and the crime severity, and compare all possible couples of criminals. When this data has been compared throughout the available years we use a human centered clustering tool to represent the outcome to the police analysts. We discuss the dif_culties in career-comparison and the speci_c distance measure we designed to cope with this kind of information.

Download Data Mining to Criminal Career Analysis.pdf

Advertisement