In order to establish a more accurate and efficient identification model of tobacco origin, a random forest pruning algorithm based on adaptive genetic algorithm (AGARFP) is proposed. According to evolution degree of groups, the proposed algorithm can adapt to different selection operators; then, by utilizing the improved adaptive genetic algorithm, random forest pruning can be conducted. The samples of five producing areas are selected to build an identification model for tobacco origin, the precision of origin identification is used as the standard to weigh the pros and cons of the algorithm. Experimental results show that the classification precision of AGARFP can be as high as 94.67%, the classification effects of AGARFP are superior to that of the comparative methods, thus the effectiveness of the proposed algorithm is demonstrated.