Sunscreen index is the primary indicator of the protection effect of sunscreen cosmetics. A handheld spectrometer was used to study the relationship between sunscreen index and reflectance spectra. Three kinds of Dingjiayi brand sunscreen cosmetics, which is SPF15, SPF20 and SPF30, were chosen as experimental material. The sunscreen cosmetics were divided into 75 samples, and 60 of them were used as calibrated samples, while the other 15 samples were used as prediction samples. The reflectance spectra data were collected by the spectrometer. The data from the wavelength range between 400 and 960 nm were processed by principal component analysis method, and the results showed that the cumulate reliabilities of PC1 and PC2 (the first two principal components) reached 91%. Then partial least square analysis method was applied to build prediction models, and the remaining 15 prediction samples were disposed by this model. The results show that the prediction correlation coefficient is 0.967 7, and the prediction precision is good. So the spectral analysis method proposed in the present paper has good performance in classification and discrimination of sunscreen cosmetics, and is a new approach to test the SPF of cosmetics.