Owing to the advances in information technology and heightened awareness regarding disaster response, many evacuation simulations have been performed by researchers in recent years. It is necessary to develop suitable disaster prevention plans or evacuation plans using data generated by such simulations. However, it is difficult to understand the simulation results in their original form because of the detailed and voluminous data generated. In this study, we focus on tensor decomposition, which is employed for analyzing multi-dimensional data, in order to analyze the evacuation simulation data, which often consists of multiple dimensions such as time and space. Tensor decomposition is applied to the movement trajectory data generated in the evacuation simulation with the objective of acquiring important disaster or evacuation patterns.