| As a discipline of studying the law of nature, the hydrology which involved numerous data processing problem greatly rely on the observation samples. A mass of flood data were directly used for parameter calibration in the past. The result is that the character of one kind of samples is concealed by large of other samples, and the character missed in the parameters. So, it is necessary to do some prepare processing before the forecast for the representative of the parameter.Clustering analysis which is one of the statistical tools was introduced to solve the problem. This is a boomingly developing technology, and it is used in vast area including data dig, statistics, learning machines, space data-base, biology and market etc. Clustering analysis help divide the resembling samples into the same type and find the potential information.According to the distributing of rain-measure station, taking account of the distribute characteristic of the river area in space and time, we use rainfall volume as the cluster index to chose forty-eight ones for next step.Compared in six method of hierarchical clustering, the average-method is the best one because it is space-conservation, monotonous, often used, and have a better effect. The value c called connection-coefficient is used to evaluate which method is better than others. The regulation is that the nearer the value close to 1, the better the method is.In the prepare processing of the samples, hierarchical clustering method in Matlab statistic tool was used to divide the data into several type which include its potential character, then the module produce the parameter in every type. At last, we used the very kind parameter to predict the flood accordingly and got better effect.In summary, the combination of the clustering analysis and flood forecast have a good effect and have vast applicable foreground in the studying of flood prediction. |
