Research On Data Mining Algorithm Based On The Meteorological Data

With the development and universalizing of the computer technology, time series data mining, the research focus and the important component of data mining, is gradually applied into every field, such as finance, telecommunication, medical treatment, meteorology and so on. But the time series data mining cannot be used straightly because of which is not a pervasive and general method. For the meteorological field, this thesis researched the method.The major contributions of this thesis are as follows:1. With the practical meteorological data, this thesis concludes the feature of the meteorological data, including periodic property, season property, random property, and temporal-spatial correlation property. For the application of meteorological problem analysis of such a region, this thesis proposes the analysis framework, which combines the clustering and tendency mining.2. Based on the feature of meteorological observation, and with the latitude and longitude information, this thesis proposes a weight-based Euler distance measure method. And at the basis, this thesis combines the traditional hierarchical clustering with the K-means method, and proposes the BIRCH clustering method based on the weighted similarity.3. According to the polynomial fitting method proposed by Savitzky and Golay, this thesis designs a seventh order smooth filter as the tool for preprocessing before the tendency mining of data series. On the basis of comparison searching and sorting, this thesis proposes a comparison-based pattern division method. At last, this thesis analyzes the tendency of each pattern after division using the abstracting method for the pattern property.