For the long-term monitoring of structure, the deformation trend changes periodically and is hard to extract. A small amount of recent data can be selected to avoid such problem. The study refers to the idea of grey system theory and provides an improved way of deformation prediction in time series analysis with a small amount of data. By cumulating the original data, the trend item is made clear and the rule of data becomes obvious. The prediction results show that the way provided by this article gives a more accurate prediction in the short term. When the prediction results have a large deviation with actual deformation, it can be believed that the trend has changed and the monitored structure may be affected.