The use of digital data analysis within law enforcement is not simple. For example, how can you predict if somebody is a terrorist? Dutch researcher Stijn Vanderlooy has developed a model that makes digital data analysis more reliable.
In recent years there has been a rapid increase in the storage of digital data within the field of law enforcement. However, this data must be analysed to extract knowledge. For example, where does the perpetrator of a crime live? How great is the chance that somebody shall commit several crimes and therefore become a repeat offender? Within law enforcement the reliability of the data is, however, vitally important. And that is where the problem lies: the available computer models are not considered to be reliable enough.
The three highly promising stages for data analysis that Vanderlooy has identified, consist of different steps. First of all the data (for example persons) are no longer divided into classes but are organised according to the likelihood that they belong in a class. This approach opens up a large number of new applications, for example tracing organised credit card fraud or drawing up suspect profiles. Subsequently, if a classification of probability is used then the quality of the computer model can be guaranteed up to a desired level. Finally, an optimal computer model is designed for the reliable classification of data in more than two classes.
Vanderlooy's research falls within the research discipline of Machine learning, an aspect of artificial intelligence that provides effective and efficient models for the analysis of data. This project was part of the NWO programme ToKeN (Access To Knowledge and its enhancement Netherlands), which focuses on fundamental problems in the interaction between a human user and knowledge and information systems.
The above story is based on materials provided by NWO (Netherlands Organization for Scientific Research). Note: Materials may be edited for content and length.
Cite This Page: