An improved quantum combination method of mass functions based on supervised learning

In Dempster-Shafer evidence theory, how to model and handle the uncertainty involved in mass functions is generally concerned by researchers. Recently, a quantum model of mass functions was proposed, in which a mass function was represented as a quantum pure state with constraints. Based on that, the authors also gave a quantum averaging operator. However, the phase parameters in the quantum model were unknown and the phase differences in the quantum averaging operator were subjectively given to 0. Considering these problems, this paper proposes an improved quantum combination method of mass functions based on supervised learning, where the optimal phase parameters are derived through supervised learning so as to acquire the phase differences in quantum averaging operator. These obtained parameters are dependent on objective data. Comparative experiments on benchmark data sets are conducted to verify the validity of the proposed method.