Consider the following scenario which iMOV will make feasible:A member of the public reports a street robbery, or it may be a burglary or violent attack. The address and crime details are recorded. A computer system integrates this information with characteristics of the locale where the crime took place and other reported crimes in the area. Dedicated crime analysts work with this material by interacting with a map that shows other crimes and known offenders in the area of the assault. This geographical information is enhanced by background information on offenders' modus operandi and a set of related analysis functions, enabling a detailed analysis of the distinctive characteristics of the offence to be carried out. The behavioural signature of this crime and any forensic information, is compared with the signature of all crimes in a number of different police databases using data and text mining techniques. The results of these comparisons lead to the proposal of a number of possible offenders and their likely residential locations. By interrogating the system the analyst alerts a local police patrol to observe an area for the criminal and to pay particular attention to three most likely suspects with a note of where they are currently living. This comparison of offences, the drawing of inferences about offenders, their residential locations and their patterns of criminal activity, all consist of generating data sets from limited information and then mining that data for productive directions to continue the investigations. The data is textual, semi-structured, does not have consistent context or parameters, and is typically partial and of limited reliability. By tying analysis of this data into geographical information systems there is enormous potential for improving many aspects of police work, including crime prevention, investigation and even preparing a case for court. The domain of police investigations therefore provides an exciting vehicle and data to develop software tools, of wider applicability, say to market research or public order management.The project thus allows:1.Flexible Data Management so that data from different sources can be effectively combined, for example, information on offences with that on offender records. These are typically stored in different forms on different systems by the police, and may use varying terms for the same concepts.2.Comparative case analysis based on data and text mining. This allows offences to be linked to a common offender. Recent research, by the applicants, on volume as well as serious crime, has demonstrated that remarkably accurate links can be made using simple parameters, higher than 80% in some cases, but to use these findings effectively an interactive system is needed. 3.Prioritisation of likely suspects. A number of published studies have shown the possibility of using offence location as a basis for identifying possible suspects' residential locations and using this as a filter for searching for and prioritising suspects.4.With added demographical and land-use information it is possible to refine the search parameters for suspects beyond those available from the general geographical information. 5.Trail and sequential analyses also allows predictive models to be developed that will allow police to anticipate routes home and the likely location of future offences. 6.To improve the power of these systems it is essential to take account of base rates of criminal activity. This will also provide a basis for moving from the coarse-grain modelling of much current crime mapping to the fine-grain necessary for dealing with the actions of individual offenders.7.The most effective implementation of this system will be with real-time data collection and rapid inference. The system will therefore be developed to ensure that such processes can be integrated in a timely and accurate manner.
|Effective start/end date||1/03/06 → 31/07/08|
- EPSRC (Engineering and Physical Sciences Research Council): £98,412.00
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.