【作者】 金涛；

【导师】 沈惠璋；

【作者基本信息】 上海交通大学 ， 管理科学与工程， 2007， 硕士

【摘要】 当前,各种突发公共事件频繁发生,每年由此造成巨大的损失和恶劣的影响。但与此同时,对应各种突发危机事件的应急决策能力和基础还比较薄弱。因此,对各种突发危机事件的监测、预报、识别、控制、预案制定,事件发生时的快速响应、决策,以及相关的计算机决策支持系统,就成为研究的热点。应付危机事件,需要进行迅速、正确的判断,制定合理的对策。危机事件及其引发的问题属于结构不良或非结构化问题,与一般事件相比,危机事件具有突发性、严重性和紧迫性的特征。因此,面向突发危机事件的应急决策也相应地区别于一般决策问题,具有如下特征:(1)决策信息水平低;(2)有效决策时限短且严格;(3)危机系统处于动态演化过程中;(4)决策主体有限理性。而应急决策对决策者素质能力要求高,决策责任重大,采用群决策支持系统对应急决策提供辅助支持是可行且必要的。本文在对突发危机及应急决策特性进行深入研究的基础上,分析了目前流行的基于预案的应急决策方法及其局限性,据此提出了一种基于范例推理的迭代式应急决策过程模型。基于范例推理(CBR)是人工智能领域中一种重要的基于知识的问题求解和学习方法。将CBR方法应用到应急决策过程中,能充分利用应急决策经验;而迭代方法满足了危机事件的动态演化特征,实现了应对式的应急决策。因此,该决策过程模型降低了决策的盲目性,有助于提高应急决策的效果和效率。本文详细阐述了面向应急决策的范例推理的实现过程及相应的算法。首先提出了危机范例的形式化表示框架;给出了基于经典粗糙集和相似粗糙集技术的范例检索方法,包括范例属性约简、加权和相似度度量等的具体算法;然后,在借鉴数据挖掘聚类技术的基础上,本文提出了相应的范例库维护方法。最后,以防汛应急决策支持系统为例,演示介绍了本文所提出的决策过程模型和相关算法技术在实际中的应用。更多还原

【Abstract】 q Recently, various kinds of crisis and public emergencies happen frequently and cause great lost. Considering the still inadequate ability of crisis management and emergency response, the research on the crisis management and emergency decision making, especially the decision support method, shows its importance.To response to the crisis and incidents, quick and correct judgment as well as rational reactive decision is necessary. Both crisis itself and corresponding emergency decision making (EDM) are unstructured problems. Compared with normal problems, crisis is differentiated as sudden, harmful and serious. Therefore, the EDM process for crisis has its unique characteristics, including (1) low quality of decision information; (2) short and strict decision time limitation; (3) the crisis system is in the process of dynamic evolution; (4) decision makers have limited senses. Moreover, EDM requires high on the knowledge and ability of decision maker. Combined the heavy responsibility for decision result, introducing group decision support system into the EDM is feasible and necessary. Initiated by researching on the characteristics of EDM for crisis, this thesis analyzes nowadays prevailed emergency response method based on preparatory plans, identifies the weaknesses and problems of this kind of method, and proposes a novel iterative decision process model based on Case Based Reasoning (CBR) for EDM. CBR is an important problem solving and learning method based on knowledge. Using CBR in EDM process can make full use of previous experience. At the same time, an iterative model will fill the requirement of the dynamically evolutionary process of crisis. Hence, the proposed model can help identifying EDM’s objectives, reducing the difficulty and blindness of EDM, and improving the effectiveness and efficiency of EDM process.This thesis elaborates the realization of proposed decision process model and related algorithm in detail. It firstly delivers a framework for the case representation. Then, based on the theory of rough set and similarity-based rough set, it describes the method for case retrieval, including algorithm for case attribute reduction, weighting, and similarity measurement. Also, it introduces the case base maintenance method using clustering technique in Data Mining. Last, a sample decision support system for flood control is presented to demonstrate the application of proposed model.更多还原