Research On Spatial-temporal Data Mining In Meteorological Data Based On GIS

As the two of the most important information technologys, Data mining and GIS are very useful in processing of meteorological data. Data mining, designed to analyse data, is good at processing large meteorological information and revealing knowledge, and can effectively solve the problem of " data rich, information poor" given by the growing weather data; GIS technology,as the realization of spatial data management method, is good at displaying meteorological data. Traditional data mining is working at relational database or transactional database, and is unable to reveal the meteorological information included in the spatial and temporal characteristics. The research of spatio-temporal data mining is very important in improving the use of massive meteorological data and weather forecasting.In this paper, the data mining technology and GIS techniques are integrated, spatio-temporal data mining based on GIS are researched, the data are calculated by three methods of spatial analysis, spatial clustering and spatio-temporal association rules.Basic concept, main technical methods and relationship of data mining, spatio-temporal data mining and GIS are introduced firstly. Secondly, the though of spatio-temporal data mining based on GIS is put, the integrated methods are discussed, the integrated framework is designed, the main technical methods are resumed, the theoretical foundation of this paper is prepared.To improve quality of data, data processing is discussed in aspects of data summary, quality control and format conversion, the data foundation for the thesis is carried out.In order to explore the potential regularity of meteorological data, spatial analysis, spatial clustering and temporal association rules are researched. Spatial analysis is not only the means of processing data missing, but also the basic data mining methods. In this paper, the data is calculated using spatial analysis methods and the best interpolation methods adapted to the northeast weather data is given. To get northeast regional climate divisions, k-means algorithm is introduced. To solve the problem of initial value, the improved k-means algorithm based on the level is introduced, which is a good way to determine K value. In order to discover the spatio-temporal relationship among the data, the spatio-temporal association rule is studied in this paper, the forecast rules of visibility is calculated. Lastly, the display software is developed in order to display the result of data mining directly.