Research On Multi-source Information Fusion Method Based On Granular Computing

In an information-rich society,multi-source data presents diversity and conflicts.The fusion methods related to granular computing mainly have the following problems: they cannot be applied to multi-source decision information system with missing markers;they do not consider the variability of the sources and apply the same thresholds and weights to all granularities.In the context of multi-source information system,this paper studies how to use the idea of granular computing to construct rough set models and design uncertainty measures to fuse multi-source decision information system,and then also discusses how multi-source information system with differences can be fused.The related work is as follows.(1)In view of the problem of there are missing markers in the multi-source decision information system,three methods for measuring uncertainty are designed from the perspective of granular computing,and a fusion algorithm based on attribute importance is proposed to realize the fusion of multi-source information system.Finally,a series of comparison experiments are designed on the KEEL dataset to compare the fusion method proposed in this paper with the traditional method,which fully proves the efficiency and effectiveness of the fusion method.(2)In view of the problem of the granularity of knowledge derived from multi-source decision information system takes the same threshold,an adaptive calculation method corresponding to the granularity threshold is proposed.Then a generalized adaptive multigranulation rough set model is constructed,and the performance of the multi-granulation model under the adoption of pessimistic and optimistic strategies is investigated.It is experimentally demonstrated that the proposed generalized adaptive multi-granulation model is more reasonable and flexible than other methods of multi-granulation and integrating multisource into a single source.(3)In view of the problem of obtaining the degree of decision support for each system in a multi-source decision information system,a method for measuring the importance of sources based on conditional entropy is defined.Then the threshold pairs of different sourcederived knowledge granularities are obtained adaptively and multiple granularities are fused by weighting.Further,the weighted adaptive multi-granulation rough set model and attribute reduction algorithm are proposed.Finally,the rationality and feasibility of the proposed model are analyzed by a case study.