Groundwater is one of the most precious natural resources, and its exploitation and utilization are expanding year by year under the influence of human activities. Groundwater resources are therefore of great significance to the sustainable development of human society. Scientific prediction of the amount of groundwater resources and the establishment of a reasonable and feasible groundwater resource management system can achieve the purpose of development and protection at the same time. At present, there are various methods to predict the amount of groundwater resources, the traditional methods are neural networks, mathematical models, etc., and the combination of time series analysis methods and traditional methods derive a new type of prediction model that makes the calculation results more suitable for engineering practice, taking Guizhou Province as an example, three methods are used to predict the amount of groundwater resources of the study area, and based on the results of the prediction, it can be obtained that: LSTM modeling method in the study area Based on the prediction results, it can be concluded that the LSTM modeling method shows higher accuracy and applicability in the prediction of underground water resources in the study area, which can provide a certain reference for the planning, development and utilization of underground water resources in the region.