Selective and incremental fusion for fuzzy and uncertain data based on probabilistic graphical model

Abstract

Active and dynamic fusion for fuzzy and uncertain data have key challenges such as high complexity and difficult to guarantee accuracy, etc. In order to resolve the challenging issues, in this article a selective and incremental data fusion approach based on probabilistic graphical model is proposed. General Bayesian networks are adopted to represent the relationship among the data and fusion result. It purposively selects the most informative and decision-relevant data for fusion based on Markov Blanket in probabilistic graphical model. Meanwhile we present a special incremental learning method for updating the fusion model to reflect the temporal changes of environment. Theoretical analysis and experimental results all demonstrate the proposed method has higher accuracy and lower time complexity than existing state-of-the-art methods.