چارچوب مدیریت اطلاعات وقوع حادثه بر اساس ادغام داده ها، داده کاوی، و تصمیم گیری چند معیاره

An effective incident information management system needs to deal with several challenges. It must support heterogeneous distributed incident data, allow decision makers (DMs) to detect anomalies and extract useful knowledge, assist DMs in evaluating the risks and selecting an appropriate alternative during an incident, and provide differentiated services to satisfy the requirements of different incident management phases. To address these challenges, this paper proposes an incident information management framework that consists of three major components. The first component is a high-level data integration module in which heterogeneous data sources are integrated and presented in a uniform format. The second component is a data mining module that uses data mining methods to identify useful patterns and presents a process to provide differentiated services for pre-incident and post-incident information management. The third component is a multi-criteria decision-making (MCDM) module that utilizes MCDM methods to assess the current situation, find the satisfactory solutions, and take appropriate responses in a timely manner. To validate the proposed framework, this paper conducts a case study on agrometeorological disasters that occurred in China between 1997 and 2001. The case study demonstrates that the combination of data mining and MCDM methods can provide objective and comprehensive assessments of incident risks.

Various types of incidents, such as the September 11 attacks, the SARS epidemic, and the 2008 Sichuan earthquake, cause huge loss of human lives and properties and highlight the need to improve our capabilities to prevent, protect against, respond to, mitigate, and recover from natural and manmade incidents [11]. Incident management activities can be divided into pre-incident and post-incident phases, which require different handling approaches since they associate with different levels of pressure, complexity, uncertainty, and response time. Pre-incident phases emphasize complete and comprehensive data analysis and decision support functions, while post-incident phases require fast and often real-time responses. Fig. 1 shows the main phases of incident management. Successful incident information management depends on a continuous, consistent, and systematic process that unifies pre-incident and post-incident phases.Emergency situations often involve data collected by multiple organizations, saved in different formats, and resided in distributed sites. In order to be effective, an incident information management system needs to deal with several challenges. First, it must present the heterogeneous distributed incident data in a standardized format for accurate and timely data access, analysis, evaluation, and dissemination. Second, the system should help DMs explore large and complex data efficiently, detect anomalies, and extract commonalities and correlations (OASIS [50]). Third, the system needs to assist DMs in evaluating the risks and selecting an appropriate solution during an incident. Fourth, the system should provide differentiated services to satisfy the requirements of different incident management phases and different types of incidents. For example, the urgency of a drought event is lower than an explosive blast threat to a school building. Although there is much literature in the area of incident information management systems, the specific challenges discussed above have not been thoroughly examined. The goal of this paper is to develop an incident information management framework that concentrates on these requirements. This conceptual framework consists of three major components: data integration, data mining, and multi-criteria decision making (MCDM). The data integration component is a set of approaches designed to act as a middleware between the underneath heterogeneous data sources and the upper intelligent analysis modules. It provides a high-importable, structuralized, and unified data interface to upper applications. The data mining component uses data mining methods to identify useful patterns and presents a process to provide differentiated services for pre-incident and post-incident information management. The MCDM component describes how MCDM methods can be applied in incident management to assess incident risks, evaluate feasible alternative solutions, and dispatch emergency resources. To validate the proposed framework, this paper conducts a case study on agrometeorological disasters that occurred in China from 1997 through 2001. The case study demonstrates that the integration of data mining and MCDM methods can offer particular advantages for incident management. The contributions of this paper are threefold. It brings together the existing literature on incident information management and identifies several important requirements for incident information management systems that have not been fully investigated. The second contribution is the development of a conceptual incident information management framework that combines data integration, data mining, and MCDM methods based on those requirements. The third contribution is the use of data mining and MCDM techniques to assess the risks of natural incidents using real-life agrometeorological disaster data. The rest of this paper is organized as follows: Section 2 reviews related works. Section 3 presents the incident information management framework, including the data integration module, the data mining module, and the MCDM module. Section 4 describes a case study that validates the incident information management framework. The last section concludes the paper.

Incident information management requires immediate and effective response from DMs under pressures and uncertainties. Technological developments in areas such as data collection, storage, computation, and pattern recognition have enabled more effective incident information collection, processing, and exploration. One of the fundamental problems in incident information management is how to integrate and analyze heterogeneous incident data and provide intelligent decision support to DMs. This study proposes a conceptual framework for incident information management to support information integration, intelligent data analysis, and multi-criteria decision making. It develops a three-level framework for incident information management, including heterogeneous data integration, data mining and MCDM, and collaboration tools. Data integration level provides a distributed heterogeneous data interface that integrates various data sources and a unified data interface that facilitates differentiated services to upper application modules. Data mining and MCDM level supports in-Departmenth data analysis, helps DMs choose the most appropriate solution from a set of feasible alternatives, and takes appropriate incident response operations in a timely and efficient manner. The third level has a graphic user interface and collaboration tools to facilitate DMs to utilize and interact with the data mining and MCDM functions without knowledge of the underlying technical details and mathematical theories. To validate the proposed framework, a case study that utilizes data mining and MCDM methods to evaluate the risk of agrometeorological disasters using Chinese agrometeorological disasters data from years 1997 to 2001 is reported. The case study applies the TOWA operator, cluster analysis, grey relational analysis, and TOPSIS procedure to rank 31 provinces according to the impact of four kinds of agrometeorological disasters on their agriculture production. The results illustrate that the combination of data mining and MCDM methods can provide objective and comprehensive assessments of incident risks. Due to the limitation of the datasets used in the case study, the data integration module discussed in the framework is not explored. The future research on the incident information management framework includes two major directions. The first direction is to instantiate the data integration module proposed by developing data fusion algorithms and collecting or simulating incident data sources that reflect the heterogeneity and variety features discussed in the framework. The second direction is to implement the differentiated response mechanism and realize an automatic transition between daily incident preparedness and emergency response in an incident informaiton management system.