Abstract: Due to the lack of accurate information on loaded car arrival, it is difficult to find the mismatch between unloading volume and station unloading capacity in time, which often leads to a passive situation of freight unloading organization. Thus, the railway loaded car arrival prediction and unloading capacity early warning system is developed by making full use of existing historical data of freight traffic and real-time loaded car tracking data. This paper introduces the architecture and functions of the system and describes the loaded car arrival prediction algorithm and the graded unloading capacity warning mechanism in detail. This system can dynamically predict the arrival time of freight cars loaded with goods of different categories to be received at all railway freight stations in the next three days. Once the system discovers the number of loaded cars predicted to be received at a station exceeding its unloading capacity, graded early warnings will be triggered immediately and adjustment tactics for freight traffic organization will also be recommended automatically to assist the dispatchers working at different levels to take corresponding measures timely in a coordinated manner so as to reduce the accumulation of loaded cars waiting for unloading and shorten the turnaround time of freight cars, thus helping promote lean management of freight car unloading and improve the level of freight traffic dispatch and command.