Abstract: In order to achieve online monitoring of the quality index in the production of goat milk powder, a generic model and an independent model for protein, fat, lactose, moisture, acidity and total ash of goat milk powder were established based on NIR spectroscopy. The six quality indexes near infrared spectral data were preprocessed in 28 different ways combined seven smoothing methods and four algorithms based on modified partial least square method(MPLS), including the general model of all goat milk powder and two independent models of Shaanxi province goat milk powder and pure sheep milk powder. It was aimed to obtain the prediction performance of the three models and the applicability of the general model to the two independent models. The results showed that the standard deviations(SEP) of the 2 independent models based on the 6 quality indicators of Shaanxi goat milk powder and pure goat milk powder were 0.043~0.412 and 0.027~0.304, and the prediction coefficient of determination(RSQ) were 0.889~0.998 and 0.977~0.998, respectively. The SEP and RSQ of the 6 indicators of the universal quantitative model were 0.034~0.732 and 0.970~0.999, respectively. They all have good predictive ability. The applicability verification results of the general model showed that the general model prediction ability of protein, fat, moisture and acidity of Shaanxi goat milk powder and the acidity of pure goat milk powder was all improved when compared with the independent model. These results confirmed that the general model and independent model were feasible for fast and non-destructive prediction of the quality indicators of goat milk powder on the market. The general model had improved the predictive ability of some indicators of the two independent models, which could realize more economic online monitoring the different goat milk powder.