A research was carried out on the prediction model of the viscosity of viscous oil mixture based on neural network .The research proved that the model of viscosity of the viscous oil mixtures established by the use of the BP-neural network method is reliable. A series of prediction models of the viscosity of viscous oil mixtures of an oil field in Sinkiang was established by BP- neural network as an example. The maximum error of the model was 4.1%. The correlation between viscosity, temperature and the rate of viscous oil and thin oil matched preferable, and proved a better fit compared with linear regression and Cragoe method.