The research of the difference measure method for risk probability distribution plays a key role in the early warning decision-making management of retail supply chain unconventional emergency. However, the common difference measure indices are established by the specific density function or distribution law of the risk probability distribution. In Knight uncertain environment, only the moments of the risk probability distribution can be obtained. This study proposes the difference moment measure method of risk probability distribution based on moment generating function and fuzzy data stream clustering for the retail supply chain unconventional emergency. The big data statistical analysis is performed on the risk assessment indices to obtain the moments of the risk probability distribution for unconventional emergency. The difference of moment generating functions for unconventional emergency risk is measured by the distance function in the real vector space of infinite dimensional moments and then the difference between the real distribution and the reference distribution of the risk probability for unconventional emergency is further measured by the moments. The main contribution of this study is that we propose a new difference measure method of risk probability distribution for unconventional emergency based on cloud model method, moment generating function theory, functional function and big data fuzzy statistics technology in Knight uncertain and big date environments, which can overcome the drawbacks of the existing difference measure methods for probability distributions.