Adaptive weighted generalized multi-granulation interval-valued decision-theoretic rough sets

With the development of information technology, the sources of information are increasing. How to make good use of information from different sources to make correct decisions is an important problem in multi-source information systems. From the perspective of information granulations, each source can be regarded as a granular structure and the importance of different granulations may be different in multi-source systems. We provide a weighted generalized multi-granulation interval-valued decision-theoretic rough set model (WGM-IVDTRS) for multi-source decision fusion. Firstly, the basic form and important properties of the WGM-IVDTRS model are studied and a granulation weighted method based on the classification accuracy of decision tree learning is proposed from the machine learning point of view. Secondly, three types of the WGM-IVDTRS model are established based on different determination methods of decision risk parameters. Finally, the WGM-IVDTRS models are first compared with multi-granulation decision models and other weighted granulation methods. Moreover, the performances of three WGM-IVDTRS models based on the classification accuracy weighted method are also compared. The experimental comparisons show that the importance, feasibility and effectiveness of the proposed WGM-IVDTRS models, and the third WGM-IVDTRS model performs best when people can accept the range scalability and fault tolerance of intervals. (C) 2019 Elsevier B.V. All rights reserved.