NIRS was used in rapid qualitative and quantitative detection for melamine of pure milk in this paper. Experiment was conducted by preparing two groups pure milk samples which melamine content is different for qualitative analysis and quantitative analysis. By combining NIRS technology with the cluster analysis method, A effective classification can be made on the two kinds of milk samples with and without melamine; To achieve this, spectrum pretreatment and wave length choice methods were employed before model optimization. The results showed that NIR models of predicting melamine content in pure milk has good stability and predictive ability.This paper suggested that NIR could be used as a quick, green and convenient method for predicting melamine content of dairy.