Dynamical Information Fusion of Multisource Incomplete Hybrid Information Systems Based on Conditional Entropy.

As the arrival of big data era, the data derived from practical applications are characterized as multi-source, heterogeneity and incompleteness. Such data are called as multisource incomplete hybrid data, which can be expressed by Multisource Incomplete Hybrid Information Systems (MIHIS). This paper focuses on dynamic maintenance of information fusion in MIHIS when the objects evolve with time. Firstly, we introduce the information fusion method of MIHIS based on conditional entropy. Then, to make the whole process of information fusion more intuitive, the methods for the computation of conditional entropy are introduced from the perspective of matrix in MIHIS, which is a critical step during the whole process of information fusion. Furthermore, the incremental mechanisms for maintaining the fusion of MIHIS are designed when adding and deleting objects. Finally, we employ an illustration for validating the availability of our presented incremental fusion strategies.