Research And Application Of Hydrological Time Series Trend Analyze

Data mining in time series explores the valuable algorithms and technologies from mass time series data, which contains the similarity research, trend analysis, periodicity analysis and mining the time period patterns. The trend analyze of time series makes great contribution to the dynamic change and trend prediction of time series.With the increasing amount of data of time series in hydrological databases, it has a significant practical importance in water resource management and flood dispatching to study the methods of trend analysis and then find the rules and tendencies contained in the hydrological time series. In the process of time series data mining, this paper transformed time-series to trend sequences combined with the hydrological knowledge. We focused on the segmentation of time series based on the featured data point, linear approximation and trend abstraction. Trend sequence could display the trend information of the time series.The work of this paper mainly includes:(1) This paper pretreated the hydrological time series data, including the statistical analysis, elimination of singularity data value, fill-in of the missing data, and the analyzed the impact to the trend analyze.(2) This paper proposed the time series segment algorithm based on the featured data point, and did linear approximation and error analysis to the segment sequence. This algorithm could avoid the impact of noise and had good capacity of data compression.(3) We explained the concepts of trend sequence, tend mapping and trend set, and extracted the trend attributes to form the trend sequence.(4) We did trend analyze to the time series, including the trend changes based on the time granularity and trend prediction based on the corresponding periods of history.(5) Combined with the hydrological knowledge, we mined the data on the real time database and made analysis and assess of the experiment results.