Research On Multi-Granularity Information Fusion Method For Multi-Source Data

With the rapid development and popularization of various high-tech technologies,various large-scale and complex data have emerged,such as data with massive,high-dimensional,multisource,and dynamic characteristics.In order to obtain the necessary information or knowledge from diverse sources of information,information fusion technology is usually used to convert multi-source information into easily understandable or useful information.Among numerous information fusion technologies,the multi-granularity fusion method is an information fusion method that directly obtains information from multiple sources of data.Compared to traditional information fusion methods,the multi granularity method can reduce the problem of information loss in the information fusion process.At present,multi granularity information fusion methods are widely applied in fields such as data mining,knowledge discovery,and uncertainty processing.This thesis mainly focuses on the dynamic update of multi-granularity fusion methods in multi-source decision information systems,as well as the research on multi-granularity fusion methods and dynamic update mechanisms in multi-source decision information systems with different source importance.The main research work is as follows:1.Considering the dynamic situation in the multi-source decision information system,a dynamic algorithm based on matrix calculation to dynamically update the multi-granularity fusion operator is designed.Firstly,the method of calculating the multi-granularity fusion operator based on the matrix is constructed,and then the dynamic update mechanism of the multigranularity fusion operator in the four dynamic situations of the multi-source decision information system is discussed and constructed.Finally,a numerical experiment is carried out using the public data set,and the effectiveness of the proposed dynamic update algorithm is verified by comparing with the static algorithm and other algorithms.2.Considering the different importance of information sources in multi-source decision information system,a multi-granularity fusion method based on weight is constructed from the perspective of the uncertain information quantity of information sources.First,the conditional entropy of different sources is defined to measure the amount of uncertain information contained in the sources,and then the source weight is determined by the conditional entropy.Then,the weighted multi-granularity information fusion method in multi-source decision information system is established by using the source weight and the characteristic function of the source.Finally,the experiment is carried out with the public data set,and compared with other traditional fusion methods.The approximate accuracy of the fusion operator shows that the proposed weighted multi-granularity method is better than other traditional fusion methods.3.Based on the weighted multi-granularity fusion method,the dynamic update mechanism of the weighted multi-granularity fusion operator in four dynamic cases of the multi-source decision information system is discussed,including the simultaneous increase of information source and attribute,the increase of information source and attribute,the decrease of information source and attribute,and the simultaneous decrease of information source and attribute.Finally,a numerical experiment of the corresponding dynamic situation is carried out using the open data set,and the dynamic algorithm is compared with the static fusion algorithm to verify the effectiveness of the proposed dynamic algorithm.