Suspicious Criminal Behavior Prediction Based on Roaming Trajectory Characteristics

Roaming around the crime scene is a characteristic of some criminals’ behaviors before committing a crime. In order to achieve the criminal purpose, it is common for criminals to check the crime scene many times in advance. However, the definition of the roaming behavior is a vague concept. This study firstly describes the relationship between the predetermined crime scenes and the roaming behaviors of suspects. Then we propose an algorithm to identify suspicious trajectories with roaming behaviors. The first step of the algorithm is to identify suspicious trajectories by applying a clustering method to extract the stagnant areas from different trajectories in real time. The second step is to identify the trajectory with loops from the suspicious trajectories based on the suspect’s roaming behaviors for casing the site. The third step is to calculate a corresponding suspicious value by using speed and timestamp of the trajectory and then combine with crime rate of different areas to design an appropriate weighted scheme. Experiments with prisoners’ trajectory data has proved the feasibility of the proposed method. Furthermore, the proposed method can effectively predict the abnormal behaviors of community inmates.