Similarity Study of Hydrological Time Series Based on Data Mining

摘要

The rapid development of data mining in recent years provides new research ideas and methods for hydrologic time series similarity, which can be used to mine hidden patterns and laws in massive hydrologic data as well as useful information for predicting hydrological processes. By using the hydro-logic time series data, through the deep learning and data mining related technology, the hydrologic time series prediction research. In order to better study the hydrological time series similarity based on data mining, this paper use the integral autoregressive moving average model (ARIMA) model for linear autocorrelation is part of the hydrological time series prediction, then use support vector regression (S VR) model to predict nonlinear part, the forecast results together and get the confidence of A confidence interval, which determine the actual value is not in the confidence interval of outliers, the last in a river basin liuhe measured data validation of the test sites, the results show that the similarity detection efficiency of 32.9%.