Abstract: In this paper, according to visible-near infrared spectral data of drugs, we proposed a method for drug detection based on support vector machine （SVM） and BP neural network regression model. Analyzing and researching on the data set of the visible-near infrared tablets released by the International Diffuse Reflection Conference （2002）. The results showed that the prediction accuracy of the SVM model was significantly higher than that of the BP neural network model, and it can be applied to the visible-near infrared spectrum detection of drugs and provides an accurate and effective method for drug testing.

Key words: support vector machine, BP neural network, visible-near infrared spectroscopy, drug detection, prediction