A Novel Multi-Source Information Fusion Method Based On Dependency Interval

Uncertainty exists everywhere in the real world.How to deal with uncertainty information efficiently is the key research object of artificial intelligence.In the era of big data,there are many forms of information,many sources of information and huge amount of information.In order to effectively mine knowledge from multiple information sources,information fusion technology is usually used to transform and fuse multi-source information systems,and finally integrate these information together.Among many information fusion technologies,rough set theory is a kind of tool that does not need any prior information to deal with imprecise,inconsistent and incomplete data.At the same time,granular computing theory is a new method to solve complex problems by decomposing complex information into many information grains,which has been widely used in dealing with uncertain data.Many information fusion methods for single-valued information system transformation are derived from granular computing theory,but the transformation,from single-valued information system to interval-valued information system,is not studied.It is well known that interval-valued data can describe the uncertainty and random change of data more effectively.Therefore,the thesis proposes a multi-source information fusion mechanism based on dependency interval.In addition,in many practical application scenarios,equipment failure or improper operation will result in abnormal data collected from multiple information sources,etc.Besides,the equipment renewal and the theory consummates make the collection information the channel to increase.If we continue to use the static fusion mechanism,it will lead to the repeated calculation of the known data and reduce the fusion efficiency.Therefore,this paper designs dynamic fusion mechanism for four cases of information source and attribute change in order to reduce the time needed to use static fusion mechanism and improve the efficiency of dynamic fusion.The main work of this paper is as follows:1.Using statistical tools to construct dependency function and dependency interval,considering the length of the generated dependency interval and the number of data points in the dependency interval to define the dependency function,the data points corresponding to the same data points of each data source are calculated,and the data points corresponding to the maximum values of the dependency functions on both sides are found as the boundary points of the generating dependency interval according to the requirements of the definition of the interval.Finally,it is integrated into an interval-valued information table,and then the multi-source information system is transformed into an interval-valued information system.Experimental results on UCI data set show that the proposed fusion method is superior to three common fusion methods in classification accuracy and quality.2.Because multi-source information system can change information sources and attributes,this paper designs and establishes four dynamic fusion mechanisms of attributes and information sources according to four dynamic changes of multi-source information system.At the same time,the corresponding dynamic fusion algorithm is designed,and their time complexity is analyzed and compared with the static fusion algorithm.Finally,the experimental results show that the dynamic fusion method is more efficient than the static fusion method.