关键词: 案例检索, 相似度, 突发事件, 舆情演化态势, 事件链

Abstract: In recent years, emergencies have caused serious loss of life and property in China and the international community, and their characteristics of uncertainty, suddenness and dynamism often force decision-makers to draw on past event-handling experiences to rapidly formulate emergency response plans. In addition, with the development of mobile Internet and smart terminal devices, once an emergency occurs, the related information derived from it usually receives a large number of comments and retweets from the public, and this kind of information interaction promotes the evolution of public opinion, which, in turn, has a counter effect on the process of handling emergencies. Therefore, it is of great significance to propose a more targeted case retrieval method by considering the influence of public opinion evolution situation on the process in the context of decision makers drawing on historical case experiences to solve current case emergency response problems.
However, the existing case retrieval-based emergency response decision-making methods seldom take into account the impact of public opinion information derived from emergencies and its evolution situation on decision-making results, and it is difficult to conduct research from the overall perspective of the event chain. To this end, the problem to be solved is how to quantify and incorporate the effects of the decision maker’s risk perception preference and the evolution situation of public opinion on the case retrieval process according to the decision-making information such as the given case attributes (including those of the event and the public opinion information) and their weights, in order to retrieve the most similar historical cases from an event chain perspective. Based on this, a two-stage emergency case retrieval method in response to emergencies with the influence of the public opinion evolution situation is proposed. The basic idea of this method is: on the one hand, based on the purpose of case retrieval, the historical cases that are most similar to the target case are retrieved as a kind of expectation of the decision maker, and each attribute value of the target case is selected as the “reference point”, and the prospect theory is applied to incorporate the decision maker’s behavioral characteristics, such as the risk perception preference; on the other hand, the formation process of the event chain is considered: when an emergency event triggers the initial public opinion information, its initial situation does not have a substantial impact on the event handling, and the case retrieval process at this time only needs to focus on the emergency event. However, with the continuous evolution of public opinion, once its influence exceeds a certain threshold, it will affect the emergency response process of emergencies or even trigger new crisis events, and the case retrieval process needs to be adjusted according to the current evolution situation of public opinion. In the research process, firstly, the heat value of public opinion information is calculated to determine the influence degree of public opinion evolution situation, and the stage of case retrieval is determined; secondly, the similarity of event and its attributes is calculated by incorporating the decision maker’s risk perception preference, and the first stage of the case retrieval process is constructed; on this basis, the comprehensive similarity of cases is obtained by calculating and assembling the similarity of public opinion information and its attributes, and the second stage of case retrieval process is constructed; finally, the feasibility and validity of the proposed method are verified through case application and method comparison.
The innovation lies in: first, the risk perception preference of decision-makers for attribute values is introduced into the calculation of case similarity; second, considering the uncertainty of the influence of public opinion evolution situation in different stages, the best similar historical case is determined by organically assembling event similarity and public opinion information similarity. The results show that the proposed method can not only reflect the important role of decision makers’ risk perception preference in emergency case retrieval, but also reflect the impact of event-derived public opinion information and its evolution situation on decision-making results. The above research results can not only lay a certain foundation for the application of behavioral decision-making theory in emergency case retrieval, but also provide feasible ideas for the study of this problem from the event chain rather than the single event level.
For future research, two directions will be considered: first, only the case attribute values of exact numbers, interval numbers and fuzzy linguistic variables are considered in this study, and other data types need to be further refined and incorporated; second, the proposed method only considers the secondary event chain composed of the event and its derived public opinion evolution situation, and in the future, secondary events that may be triggered by public opinion can be further introduced into the emergency case retrieval process, so as to propose a retrieval method applicable to the tertiary event chain.

Key words: case retrieval, similarity, emergencies, public opinion evolution situation, event chain