The purpose of this study was to investigate taxi driving cycle and emission factors in Haikou. Through the collection of taxi driving data in Haikou City, a short-stroke was used to extract kinematic fragments from the driving data, and a series of characteristic parameters were used to characterize the driving modes of each segment. The SOM clustering algorithm was used to cluster the kinematic fragments, and the data fragments were extracted based on the cluster analysis, and the taxi driving cycle in Haikou were constructed. The Pearson correlation coefficient was used to verify the correlation of the constructed driving cycle. This method can restore the typical traffic conditions under complex traffic conditions. Based on the data of taxi distribution and driving cycle in Haikou City, the COPERT model was used to calculate the taxi emission factor in Haikou City. Compared with other driving cycles already existing in the world, the taxi driving conditions in Haikou City are characterized by short idle time, frequent acceleration and deceleration, high average acceleration and deceleration, and long-term slow-moving mode. The CO, CO2, VOC, and NOx emission factors for taxis in Haikou are 0.329, 2.230, 0.170, and 0.362 g/km, respectively.