A rapid identification method for geographical traceability of dark tea based on the miniature near-infrared spectroscopy combined with chemometrics was proposed to provide transparent and reliable traceability information about the areas of producing dark tea， standardize the order of tea market and protect rights of consumers. The miniature near-infrared spectrometer was used to collect spectral information in situ of dark tea including the An-tea in Anhui Province， Tibetan tea in Ya'an City of Sichuan Province， Anhua dark tea in Hunan Province， and Chin-brick tea in Hubei Province. The simulated annealing algorithms， particle swarm optimization algorithms， ant colony optimization（ACO） algorithm， and iterative variable set optimization method combined with linear discriminant analysis and extreme learning machine（ELM） method were used to construct a geographical traceability classification model for dark tea. The results showed that the accuracy of the ELM prediction model based on the 10 characteristic wavelength information related to the geographical traceability of dark tea extracted with ACO was 97.5%. It is indicated that the combination of the miniature near-infrared spectroscopy and chemometric methods has good practical value for the geographical traceability of dark tea from major production areas in China.