Abstract: Starch, proteins and total flavonoids were the vital nutrients and pharmacological function factor in buckwheat, to achieve the fast detection of the buckwheat internal composition had an important theoretical significance and application value for buckwheat breeding, processing and other steps.Forty buckwheat samples from different origins had been selected. The starch, proteins and total flavonoids content of buckwheat were determined, and the near- infrared spectrum of buckwheat had been obtained using FieldSpec 3 near- infrared spectroscopy. Forty samples were randomly divided into the prediction set and validation set with 30 and 10samples respectively. Compared with different pre- processing methods, the prediction models of buckwheat starch proteins and total flavonoids content had been established using the combination method of principal component analysis and artificial neural network, finally the models had been verified.The results showed that the correlation coefficient between the prediction value and measurement value of buckwheat starch content was0.8719, and the relative error was smaller and its mean value was 1.82%, the method of the buckwheat starch content prediction was feasible.But the prediction for buckwheat protein and total flavonoids content was not ideal, need to be further studied.