The water content and activity of rice samples were determined by direct drying method and water activity meter, respectively. The measured values are considered as reference values. Furthermore, rice samples were measured by LF-NMR, and the transverse relaxation data of the samples were generated. A multivariate calibration model was created by using chemometrics algorithm to rapidly determine the water content and activity of rice. PLS and BP-ANN methods were used to train 160 calibration set samples and establish the multivariate calibration models, respectively. 90 prediction set samples were predicted by the calibration models. The results show the correlation coefficients between the predicted and reference values of the moisture content of the predicted set samples for the PLS and BP-ANN methods were 0.937 6 and 0.955 5, respectively, and the predicted root mean square deviations were 0.005 8 and 0.004 6, respectively. The correlation coefficients between the water activity prediction value and the reference value for the PLS and BP-ANN methods were 0.983 0 and 0.993 4, respectively, and the predicted root mean square deviations were 0.009 2 and 0.006 2, respectively. The results showed that both methods can quickly and accurately predict the moisture content and activity of rice.