User Preference-Based Knowledge Acquisition For Emergency Events

With the improvement of our emphasis on emergencies and degree of informationization, large amounts of raw data in emergencies are accumulated, which contain much valuable knowledge and rules to be extracted. But the process of data collection exists randomness and blindness caused by natural factors and human factors, which hinders the discovery of knowledge and laws of emergency. Existing researches on knowledge acquisition mainly focus on the objectivity and conciseness of knowledge, neglecting the preference of users. It makes the results not practical to users and emergency decision-making supporting. Thus, this paper proposes a method of knowledge acquisition considering user preference. It mainly contains three procedures:preprocessing, attribute reduction and rule extraction.Firstly, preprocessing of knowledge acquisition is conducted. The characteristics of emergency event and its data are analyzed. The preference of emergency deciders to knowledge acquisition is defined and quantified, including preference of attribute selection, attribute granularity and rule confidence. The decision table is constructed by recognizing the decision attribute. According to the characteristics of emergency data, the preprocessing method to deal with missing data, redundant data, noising data, hybrid type and granularity data is proposed. The decision table is coded to improve the efficiency.Secondly, a method of attribute reduction for emergency events is proposed, considering the characteristics of emergency events and user preference. It finds an attribute reduction through a heuristic algorithm by adding attributes, which combines attribute significance with user preference of attribute selection to define heuristic information. To simplify the procedure of computing, a method to measure attribute significance is proposed based on OWA operator. It computes the attribute significance of a single object and aggregates them as the comprehensive attribute significance.Thirdly, to solve the problem of knowledge granularity and practicality in knowledge acquisition, a method of hierarchical rule extraction based on user preference is proposed. It starts from the most abstract general decision table, and drill down by choosing an attribute to refine according to its significance and user attribute granularity preference. A modified artificial bee colony algorithm is proposed to extract rules from a certain general decision table. In addition, the rule extraction method uses user preference of rule confidence as the algorithm termination condition to control the granularity of knowledge.This paper aims to discover potential knowledge of emergency events based on rough set theory. It involves user preference and idea of concept hierarchy, and solves the problem of information loss, hybrid attributes, granularity and practicality of rules in knowledge acquisition of emergency events. This method can not only derive concise knowledge, but also satisfying user preference. It contributes a lot to support the emergency decision-making and disaster prevention and mitigation, and has certain theoretical and practical significance.