Moisture content is an important index in the drying process of Hami big jujubes which has an important influence on its appearance, taste, storage and transportation. Therefore, in order to realize the accurate prediction of the moisture content of Hami big jujubes, GA-ELM prediction model of the moisture content of dried Hami big jujubes was studied by using Near-Infrared spectroscopy combined with variable preferred method. In order to improve the stability and prediction accuracy of the model, the effects of kernel function and the number of neurons on the GA-ELM prediction model were discussed. Various pretreatment methods were used to deal with the spectrum of the whole band. The comparison analysis denoted that the standard normal variation (SNV) method was the best. The characteristic wavelengths were screened from the range of 927.77~2 501.14 nm by combining with successive projection algorithm (SPA), the synergy interval partial least squares (si-PLS, genetic algorithm (GA) and their combination algorithms after processing of SNV. Respectively, the corresponding GA-ELM prediction model was established. The GA-ELM model with 14 characteristic wavelengths screened by SNV and SPA had the best effect while compared with the full-band GA-ELM model. Furthermore, the predicted results could be given as follows: Rc and Rp are 0.984 2 and 0.967 5, RMSEC and RMSEP are 0.006 1 and 0.007 9 while RPD is 3.678 8. The results denoted that the SNA+SPA+GA-ELM method can realize the accurate prediction of moisture content of dried Hami big jujubes and provide a reference for the application of near-infrared spectroscopy in the on-line detection of dried Hami big jujubes.