Anthocyanin in purple sweet potato is easy to degrade due to environmental factors during the storage, resulting in changes of purple sweet potato color and reduction of nutritional quality. The changes of anthocyanin in purple sweet potato during storage were analyzed by using nearinfrared spectroscopy, and rapid and nondestructive testing model was established. The near-infrared spectra of 120 purple sweet potato samples were collected at different storage times (0d, 2d, 4d, 6d, 8d, 10d, 12d, 14d, 16d, 18d, 22d and 30d). In the spectral region between 4000cm-1 and 10000cm-1, statistical mathematical models of anthocyanins in purple sweet potato were established with different preprocessing methods (Savitzky-Golay, first derivation and standard normal variate) using the partial least squares (PLS, SNV-PLS, iPLS and GA-PLS) methods. The results showed that standard normal variate (SNV) transformation was the best preprocessing approach. The iPLS and GA methods were used to select characteristic wavelengths, and the GA-PLS model was the best among the models developed, the R2v and root mean square error of validation values were 0.9136 and 7.2398mg/(100g)，the residual predictive deviation value was 3.3397. The optimal prediction model was verified, where the R2p and root mean square error were 0.8314 and 10.7663mg/(100g), respectively, and the residual sum was -10.0417mg/(100g). These results confirmed the feasibility of using near-infrared spectroscopy for the nondestructive detection of anthocyanin of purple sweet potato, which can provide a reliable method for intelligent screening of raw materials of purple sweet potato and quality monitoring during storage.