In order to make radar target track simulation more realistic in simulation test-bed of C3I system, a new method which utilizes data mining techniques based on historical radar data is proposed. This method includes two stages: The first is “Filtering separation – piecewise fitting – feature clustering”. In this stage, the radar historical data is divided into the actual true track and noise. Through computing the second-order discrete curvature, the actual true track is decomposed into several segments, such as straight line and arc, which are fitted with multinomial subsequently.