The association rule algorithm was introduced into the prediction of seepage behavior of earth-rock dams, and a prediction model was established. Firstly, the seepage flow, upstream water level, rainfall, downstream water level, etc. were formed into a transaction group. Secondly, the K-means algorithm was used to preprocess the measured data, and then the Apriori algorithm was used to mine the frequent item sets, resulting in strong association rules. Finally, strong association rules were used for the prediction of the seepage flow. This model can better explain the intrinsic reasons of the seepage variation. The prediction model was applied to a certain earth-rock dam project, and the results show that the model has good accuracy.