KDD - Overcoming Massive Data Streams for Intelligence Tasks
Abstract
Actually very interesting IT systems promise to reveal connections between apparently harmless and unrelated information pieces. An article from the New York Times in February 2006 makes clear that common data mining techniques were not successful in general. Despite huge investments, correlating data from different sources did not yield satisfactory results. Transforming low-level data by aggregation to meaningful events is nevertheless the key to building the basis for succeeding decisions in the context of situation reports. More realistic and manageable is an approach that includes interactions with the user along with domain specific knowledge. Gaining security relevant messages should be based on an iterative multi-level process. This process represents the core element of intelligence analysis systems which play an important role for supporting decisions in management information systems. The following example illustrates the principal automated process for discovering communication structures in the context of radio reconnaissance: A crucial part of this process is the analysis and visualisation of communication structures, or more generally, of network information. This should be embedded in spatio-temporal data analysis with geo-oriented data access and the integration of domain-specific analysis functions. The intelligent analysis of radio emission data is based on data mining techniques, cluster visualisations to validate the results, a model based communication detection (including domain-specific knowledge) and the visualisation of communications. The following use case of a simple simplex communication clarifies the problems and the applied methods. Module coupling is realised by a distributed architecture. Given are a huge amount of radio emissions which are arbitrarily distributed. Each emission is described by the attributes ID, frequency, modulation type, starting time, end time, latitude and longitude.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 2006
- Accession Number
- ADA477157
Entities
People
- Vera Kamp