A framework for developing intelligent information systems to support decision making in complex and uncertain environments
Date
2009-01Author
Mateou, Nicos H.Publisher
Πανεπιστήμιο Κύπρου, Σχολή Θετικών και Εφαρμοσμένων Επιστημών / University of Cyprus, Faculty of Pure and Applied SciencesPlace of publication
ΚύπροςCyprus
Google Scholar check
Keyword(s):
Metadata
Show full item recordAbstract
Πολλές υπάρχουσες μέθοδοι ομαδοποίησης αντικειμένων/κόμβων σε γράφους ιδιοτήτων θεωρούν ότι οι ιδιότητες των αντικειμένων είναι το ίδιο σημαντικές ή αγνοούν την ύπαρξη συνδέσεων πολλαπλών τύπων. Επίσης, ανακαλύπτουν ομάδες που χαρακτηρίζονται από ομοιογένεια χαρακτηριστικών και είναι πυκνά συνδεδεμένες (densely connected components). Ωστόσο, η αναγνώριση ομάδων αντικειμένων που μοιράζονται παρόμοιες συνδέσεις είναι επίσης σημαντική. The main goal of this research is the development of an innovative Intelligent Decision Support System within the framework of Genetically Evolved Fuzzy Cognitive Maps. It is intended for the modelling of real world problems and the supporting of decision making processes. Specifically, the proposed framework is based on encoding experts’ assessments on the nature of a crisis under consideration. This assessment is then inputted in a fuzzy knowledge base that uses a linguistic form which it is modelled and processed by means of a Fuzzy Cognitive Map (FCM). FCMs are an alternative approach to decision making processes as they expand the capabilities of Decision Support Systems (DSS) and Expert Systems (ES) and also support scenario analysis and forecasting. During this research, several drawbacks of FCMs were identified and addressed. The first involves the invariability of the weights that participate in the configuration of a given problem. The second lies within the inability of the method to model a certain situation by performing all possible computational simulations following the change of a certain weight or group of weights. We addressed this issue by combining FCMs with Genetic Algorithms (GA), thus creating Hybrid Evolutionary Fuzzy Cognitive Map models. Another two important improvements to the FCM theory were also proposed in the present thesis. The first concerns the handling of the “Limit Cycle phenomenon” attempting to improve the inference procedure, while the second improvement refers to the use of a new structured approach named Multilayered Fuzzy Cognitive Maps for the development of FCM-based systems that are able to handle large-scale, complex systems. The methodology was successfully applied in practice where several real world problems were modelled using the proposed framework, based mostly on the fields of crisis management, political decision-making and strategy definition. The effectiveness and reliability of the method proposed has been demonstrated by means of case studies in a number of problems related to the Cyprus issue. The application of the proposed methodology is not limited only to political or crisis management problems but can be further extended, without any restrictions, to other domains due to its generic nature and simple and straightforward steps.