We have conducted long-term structural safety monitoring at the central tower of the Bayon Temple which was built in the 12th century, in Angkor monument, Cambodia. In addition to long-term monitoring of displacement at cracks, environmental measurements such as temperature and rainfall have also been carried out. The purpose is to analyze the influence of environmental factors such as temperature and rainfall on the displacement of cracks. In this paper, we use these monitoring data to develop an analysis method using Nonlinear Auto-Regressive eXogenous (NARX) neural network for the displacement of cracks in the site, which is judged to be important for evaluating the long-term structural safety of the dry masonry tower. In NARX used in this paper, the displacement at the crack is used as the output, and environmental data such as rainfall and temperature are used as the input. NARX predicts current or future displacements using models learned from past data. We verify the accuracy of these predictions using multiple parameters and models.