A systematic risk analysis for mitigation purposes plays a crucial role in the context of emergency management in modern societies. It supports the planning of the general preparedness of the rescue forces and thus enhances public safety. This study applies the principles of knowledge discovery and data mining to support the development of a risk model for fire and rescue services. Domestic fires, which are a serious threat in an urban environment, are selected to demonstrate the methods. The aim of the research is to identify important factors that contribute to the probability of the occurrence of domestic fires. Various physical and socio-economic conditions in the background environment are analysed to provide an insight into the distribution of domestic fires in relation to underlying factors. Following the cross-disciplinary nature of data mining, this study offers a set of distinct methods that share the same goal - to identify patterns and relationships in data. The methods originate in different scientific fields, such as information visualisation, statistics, or artificial intelligence. Each of them reveals different aspects of the existing relations, which supports an understanding of the phenomenon and thus expands the expert knowledge. The application of data mining techniques is not straightforward because of the specific nature of geospatial data. This study documents the analysis process in order to provide guidelines for potential future users. It considers the suitability of the methods to handle spatial and spatio-temporal data with special attention to the GIS-motivated conceptualisation of the problem being analysed. Furthermore, the requirements for the user to be able to apply the methods successfully are discussed, as is the available software support.
展开▼
