To effectively grasp the dynamic information of key regional population aggregation, guarantee timely dispatching of regional population, and prevent group security accident, this paper took Guangzhou Railway Station hub area as an example, and obtained real-time statistics of its regional population by carrying out information processing based on massive mobile phone signaling data and mapping the data to the study area using geographic information system. Meanwhile, it analyzed the change of the population in this area during Spring Festival, and summarized the periodic variation characteristics of the regional population. Based on this, this paper constructed the k value adaptive calculation model with minimum absolute percentage error, and built the k-Nearest Neighborhood algorithm based on the mobile communication data to predict urban hub traffic. The algorithm was tested under the traffic conditions on weekends and weekdays near Guangzhou Railway Station. The results showed that the average absolute percentage error (MAPE) of the prediction algorithm was around 5%. It is more accurate to effectively predict regional population.