For the dam seepage safety analysis and early warning，this paper adopts the K-mean and local outlier coefficient method.By analyzing the temporal and spatial distribution and evolution law of LOF value of dam seepage monitoring data，it reveals that the seepage time series perform scale invariance，which can effectively test local and regional outliers，and quantitatively judge the abnormality of monitoring data.Combined with the engineering case，the LOF value is calculated by MATLAB program，and the early warning threshold of dam seepage damage is determined based on statistical method，which provides a technical reference for monitoring and early warning.