Research of rainfall weather model based on multi-dimensional time series data mining
Han Hai-tao · Jisuanji gongcheng yu sheji · 2010
To research the relationship between rainfall and related meteorological factors and search for regularity of variety for those factors before and after rainfall,a multi-dimensional time series data mining model is presented.Firstly,in order to eliminate irrelevant redundant dimensions,the model performs dimensions selection for time series of meteorological factors.Then,a new extremum slope piecewise linear fitting method is introduced to implement multi-dimensional meteorological time series segmentation,data compression and feature value extraction.Finally,k-means clustering method is used to symbolize processed multi-dimensional series,and the rainfall model is extracted by rules extraction.The results of experiment show this model has a great practicability.