In order to quickly and accurately detect the status of paddy during storage, the near infrared (NIR) technique was used to establish classification models for identifying the storage quality of paddy. The NIR data of 285 samples in the range of 1 000～1 800 nm were collected. Based on the measured fatty acid, the storage state of the sample was divided into three categories (good storage quality, moderate storage quality obviously tending to decline, and poor storage quality). The optimal 10 wavelengths were selected by neighborhood rough set (NRS) algorithm. The best classification model was established by the combination of NRS and random forest (RF) algorithm. The correct classification rate (CCR) of the calibration set and the test set are 96.31% and 93.68%, respectively. The results of sensitivity and specificity are distributed in a range of 0.93 to 0.99. Furthermore, the performance of the model is also superior to the other models established by using successive projections algorithm (SPA) and principal component analysis (PCA) algorithms combined with RF. The result indicated that the fusion of NIR technique and the NRS and RF algorithms is feasible for the identification of paddy storage quality, and which can provide a reference for the development of on-site rapid inspection equipment for grain quality and safety.