Expert Finding And Opinion Fusion Oriented To Emergency Decision

The doctoral dissertation named "expert finding and opinion fusion oriented to emergency decision" is supported by the national social science foundation of China (Grant No.13&ZD173) and the fundamental research funds for the central universities (Grant No.2014104010202). This paper consists of eight chapters, including the introduction, the related basic theories, the research idea, the body from the 3rd chapter to the 6th chapter and the conclusion and prospect. As core sections, ranking fusion, feature fusion, effect analysis of experts team construction for emergency decision-making and opinion fusion together consititute the main part in sequence. In order to provide strong support for the government and other related agencies in emergency decision-making activities, this article aims to explore the construction methods and effects on expert team for decision-making consultation, and further assesses the procedures of depth fusion on expert opinion.The content obout every chapter is as follows:The feasibility of the topic has been proved from different perspectives in introduction. Firstly, combined with the reality needs and discipline attributes, author points out relative research background and illustrates research significance in theory and practice from three prospects of the national policy, the trends of multi-sources information fusion and the value of team consultation. Secondly, focused on the integration between information management and emergency services, expert ranking fusion, expert feature fusion, credibility evaluation of expert finding and expert opinion fusion, this chapter systematically reviews and summarizes the previous research. Finally, this chapter also details the key elements of the study programme, which cover research content, research methods, research emphases, research difficulties and research innovations.In the 1 st chapter, related theory closely to research content, including information fusion theory, information management for emergency, group decision theory and knowledge engineering theory, have been described. Information fusion theory has been applied to all core content, so the authors have showed its development course, the concept, fusion level, fusion method, application value and application prospect.Emergency information management theory covers information performance form in the emergency management, emergency information management process and performance assessment. Group decision theory constitutes the basis of opinion fusion, so this chapter has respectively depicted their development course, decision-making models, dynamic group decision. As an important direction for future extension, knowledge engineering has closely related with expert system building, so this chapter has highlighted its definition, the difference from knowledge management,the workflow and the application in emergency information service.Focusing on research idea,the author has firstly analysed the concept and feature of emergency decision, method for emergency decision, emergency expert consultation and emergency information need on multiple decision-makers;Secondly,to solve the problems in the activities of emergency expert consultation,the accurate and efficient process of expert finding based on data fusion has been designed,which includes expert ranking fusion,expert feature fusion and team-building effect analysis for emergency decision-making consultation;Finally,the author has outlined the idea of emergency expert fusion.Firstly, the 3rd chapter successtively makes the statement about the scene extraction of emergency and retrieval need expression. Secondly, based on previous research on information fusion for multi-sources expert features, this paper has designed reference indicators of different facets, by which expert faceted rankings are obtained. Finaly, taking the conflict given rise to by varied rankings into account, the authors have applied D-S theory and information entropy to expert faceted ranking fusion algorithm, which focuses on the calculation and synthesis of mass values.In the view of working process of sensor, the author firstly brings out three expert feature recognition methods based on knowledge resources,web resources and social network resources in sequence, in particular, this chapter gives more attention to feature recognition method of niche expert in the social media environment, in which the author takes advantage of user activity data to construct social network and does statistics of node network structure index, such as betweenness centrality, clustering coefficient. Feature and role of node in different period is distinguished via the combination of cluster analysis and time series analysis. Via comparative analysis of network statistics indexes of different clusters, the niche experts'collection can be obtained. Finally, focusing on resource balancing degree, the author has designed the method of expert feature recognition based on multiple-sensor information to solve the conflict which three obtained eigenvectors give rise to.The 5th chapter has assessed the construction effect of team for emergency decision-making.According to building processes, evaluation is divided into four parts: expert finding credibility evaluation, effect analysis of niche expert recognition, effect analysis of expert ranking fusion and effect analysis of expert feature fusion. Expert finding credibility evaluation refers to the retrieval principles and assumptions from binary indepence model, including the front-end evaluation mechanism corresponding to information retrieval and the back-end evaluation mechanism corresponding to information organization.The front-end mechanism attempts to reduce the noise in the expert feature recognition via finding the best length of expert eigenvector, the back-end mechanism deeply integrate users into the retrieval by setting user relevant feedback as the necessary reference of path selection. Taking MetaFilter as data source, this part has used relationship index and property index to illusatrate the effect of niche expert recognition via cluster analysis and time series analysis. Making the selection process of emergency expert from different areas as examples and expert ranking based singal facet as control group, outcome has demonstrated that ranking fusion method in this paper is more efficient than single method by the evaluation of ranking values in repetition and consistency.Weighted average method has been used to integrate content features of top N experts required in emergency decision, matching the obtained expert feature based on Web resources, the author has found that the degree of similarity is more than fifty-four percent, which can be accepted in similar methods.In 6th chapter, the expression and collection of expert opinion, the properties design of emergency programme, the weight design of expert assessment opinion, emergency plan assessment, the feedback of expert opinion and the example analysis of opinion fusion have been involved. Firstly, according to prior set of qualitative description language on experts assessment views, experts make assessment on emergency plan, and then three dimensional matrix including expert, property and emergency plan has been submited to convenor after the assessment views have quantitatively been dealed with via bipolar scaling method and normalization method.Secondly, information entropy has been used to calculate the weight of emergency plan properties and expert ranking mass value has been integrated to the weight of expert assessment views. Thirdly, convenor has assessed the opinions experts submitted,which includes three steps:the first step is to obtain fuzzy positive ideal solution and fuzzy negative ideal solution;the second step to calculate the distance between emergency plan and two fuzzy ideal solutions through weighted euclidean distance;the last step is to calculate membership degree of every emergency plan to fuzzy positive ideal solution, which is the standard of emergency plan ranking.Fourth, the feedback on ranking situation and analysis results have been given to every expert, experts adjust their evaluation views, and then second and third stage task have been repeated until eventual ranking of emergency plan has been received. Finally, Instance analysis about selection of forest fire monitoring sites has demonstrated that the procedure is feasible and reliable.In 7th chapter, the author has systematically summed up the expert finding and the steps of expert opinion fusion, the problems and shortage, meaningwhile, have also been pointed out, so the development of supporting data for emengency decision-making, the collection and screening of controlled data used for evaluation, the regulation and control over convenor operation need to be put on more emphasis in further research.