Modified Bioinspired Method for Decision-Making Support for Prevention and Elimination of the Emergencie's Consequences

Abstract

The work is devoted to solving the scientific problem of decision support for the prevention and elimination of emergency situations (ES) consequences based on fuzzy logic and machine learning methods. The relevance of this problem is due to the need to optimize the risk of adverse effects on human health and the environment in connection with emergency situations. The increased complexity of the tasks solved within the framework of the designated scientific problem is associated with the presence of information uncertainty in the complex accounting of heterogeneous characteristics, which in some cases can’t be normalized and brought to a single measurement scale. In such conditions, information processes for predicting the occurrence of potentially dangerous chains of events in the technogenic and natural spheres must be built using artificial intelligence methods and fuzzy logic to increase the efficiency of choosing the sequence of actions performed on the available information in order to build the necessary models, methods and algorithms to eliminate negative development situations and ensure monitoring of emergencies potential development cases. The authors give formalized statements of the tasks to be solved. A conceptual data model is proposed for constructing fuzzy decision support rules for the prevention and elimination of emergencies’ consequences. One of the options for formalizing such a data model is the transition to a vector representation of the information space. This will allow in the future to solve the problem of their classification on a set of information elements for distribution by classes of emergency situations. The criterion for evaluating belonging to a certain class is the argument for minimizing the distance between information elements in the vector space. The procedure for the accumulation by an intelligent system of precedent models set for the prevention or elimination of the emergencies consequences, which is a stage of machine learning, is described. After passing it, the intelligent system becomes capable of assessing the semantic similarity of operationally obtained models with precedents that have already fallen into the category of templates. The criterion for evaluating the effectiveness of an intelligent decision support system is the semantic similarity of the operational situational emergency model and the precedent model. A heuristic algorithm for determining semantic proximity was proposed. In order to optimize the time spent on supporting decision-making on the prevention and elimination of the emergency situations consequences, the authors also propose to use decentralized bioinspired methods, the advantages of which are internal procedures that provide diversification of the search space to exit from local optima and quickly obtain quasi-optimal solutions to the problem. The development of a modified bacterial optimization method (MMBO) was described. A software application has been created to conduct a computational experiment. The results of the conducted studies confirmed the advantages of the bacterial optimization proposed modified method.