The Research On Event Awareness Oriented Emergency Decision Making Engine

Emergency management is an important topic for governments. It provides decision makers with information, experience and knowledge support related to current emergency events. Therefore, it plays an important role in effective and efficient emergency response. Based on hard computing, traditional DSS can meet need of decision making under certainty. However, emergency decision making is usually under uncertainty, in addition, information needed for decision making has the characteristics of miscellaneous, heterogeneous, incomplete and multi-granularity. All these aspects result in difficulty in dealing with events through traditional approaches. To resolve this problem, decision making engine should have the ability of understanding and recommending uncertain information and responding to uncertain scenarios. At present, research on emergency decision making is still in exploration, no sophisticated theories and technologies have been achieved at present.This dissertation performed a deep research on knowledge support for emergency decision making under uncertainty. We focused on emergency knowledge representation and emergency response approaches. Following work has been achieved:1. Rough description logic (Rough-SHOIN) based on classic DL was proposed to be the basis of emergency response knowledge representation. Rough similarity relation was introduced into Rough-SHOIN to define rough upper and lower approximation of concepts and to extend ALC's ability in representing uncertain concepts, especially incomplete ones. Furthermore, context was considered in semantic interpretation of Rough-SHOIN to give different semantic descriptions according to context. To resolve the problem of spatial knowledge representation in emergency decision making, we extended the concrete domain of Rough-SHOIN and designed Rough-SHOIN (S). Firstly, Rough-SHOIN (S) extended spatial topological relations, and redefined RCC-8 under rough regions. Secondly, Rough-SHOIN (S) considered height relation and reachable relation which could build up 3-dimensional spatial representation framework. Rough-SHOIN (S) also included spatial orientation relation to expand the capability of oriental representation and reasoning.2. ROWL language, expanding classic OWL language, was designed to describe and store rough ontology based on Rough-SHOIN. Next, EDOM (Emergency Decision Ontology Model) was built to form the emergency decision knowledge architecture. EDOM contains upper ontology and application ontology. The former is based on eABC ontology which succeeded from ABC ontology, and the latter is a 5-tuple consisted of concepts, relations, functions, axioms and instances which described knowledge of decision making entities, emergency objects, emergency event processes, emergency event response processes and so on.3. In this dissertation, EADM (Event Awareness-oriented Decision Model) was proposed. Firstly, a formal model of emergency event awareness was built and an OSMA (Onto-based Situation Matching Algorithm) was designed. Secondly, an onto-based decision knowledge retrieving model was proposed to find and integrate heterogeneous information. Finally, an information recommender model for decision making was designed which used event information apperceived by OSMA to compute the similarity between current emergency events and historical cases, and then ranked the candidate cases aiming at different roles of decision makers, so that recommending different granularity cases to different users can be realized. The experiments showed that algorithm combining situation matching and process matching can effectively improve precision of knowledge recommendation for decision making.4. Finally, an emergency decision making engine based on above research was introduced to verify our results.