Commercially, cellulose products are designated with viscosity grade measured at 2% w/v concentration in water at 20 degrees C using an Ubbelohde viscometer. To represent viscosity/concentration curves, linear function of the eighth root of dynamic viscosity and the concentration is generally used. In this work, the influence on viscosity of aqueous solutions of methylcellulose 400 and hypromellose 4000 by temperature and polymer concentration was modelled using an empirically proposed multiple linear regression in which the transformation of viscosity by logarithm, the reciprocal value of the absolute temperature, and the concentration by square root was recommended. Due to this, the viscosity of both cellulose derivatives investigated could be predicted simultaneously with the mean difference between the observed data and the ones estimated equal to 16.2%. Expanding the linear regression with the linear interaction between logarithm of the polymer viscosity grade and square root of the polymer concentration, the precision of the viscosity prediction increased to the acceptable level of 4.1%. Other interactions between the studied variables did not provide significantly better results. The optimized regression equation enabled the prediction of kinematic, dynamic, relative, and specific viscosity of the aqueous solutions of cellulose derivatives. The dimensionless relative viscosity could be recommended because it takes into account the water viscosity at the same experimental temperature. Selecting viscosity grade of the cellulose derivative and temperature of measurement, the partial regression equations were obtained from which the relative viscosity could be determined as the function of the polymer concentration with the precision in range of 1.3-4.7%.