Scale Combinations In Inconsistent Generalized Multi-Scale Decision Systems

In the era of big data,how to conduct effective data mining models from large scale data has become a difficult problem for human beings.Granular computing,as an important scientific tool in data mining,simulates the human being's thinking,taking garnules as the basic computational unit,to establish an effective calculation model and information processing for complex data as the goal.By granulating the information with"granules" instead of "samples" to resolve the problem,granular computing has demonstrated its unique computing advantages and has aroused widespread concern among researchers in the field of intelligent information.Rough set approach plays an important role to popularize granular computing.The data description structure which is used in rough set data analysis is called an information system.Rough set data analysis does not need to provide any priori assumption beyond the date.It obtains knowledge reduction by removing superfluous information to maintain classification ability and then induces concise decision rules from decision tables.An information system in traditional rough set data analysis is usually a single-scale information table,that is,each object under each attribute can only take on one value.However,in most practical situations,the problem needs to be handled in a multi-scale environment.This dissertation mainly studies the knowledge acquisition problem for generalized multi-scale decision systems.Some basic notions related to rough set data analysis such as information systems and approximation sets,belief structures and belief functions,decision systems and decision rules are first reviewed.Concepts of generalized multi-scale information tables and scale combinations are then introduced,and definitions and properties of approximation sets under different scale combinations are further introduced.Optimal scale combination selection methods including global and local optimal scale combination selections,for consistent and inconsistent generalized multi-scale decision systems are proposed.Finally,with reference to discernibility matrix method,attribute reduction in the decision system which correspond to the selected optimal scale combination are presented,and knowledge acquisition in the sense of rule induction in consistent and inconsistent generalized multi-scale decision systems are explored.