Research Of Spatial Data Mining Approaches Based On Wavelet Analysis

With the development of science and technology, knowledge of human about the world is deepening gradually. During the course of exploring the world, all kinds of information and data come forth. And among this information, there are a large number of rules and patterns. For the purpose of finding these rules and patterns data mining is born that it is helpful to production and living of people. With the widespread application of GPS, RS, GIS, Satellite, Rader and TV camera, traditional data is gradually replaced by spatial data that it becomes the principal part of contemporary information. Consequently, spatial data mining turn into one of high lights of the current research.There are several features such as spatial autocorrelation, multi-scales and high dimensions in spatial data, which it is different with traditional data, as a result, traditional approaches does not work well for spatial data. While wavelet analysis has many favorable properties, such as frequency and time domain analysis, multi-resolution analysis, vanishing moments , hierarchical decomposition structure, de-correlated coefficients, a wide variety of basis functions and linear time and space complexity of the transformations, which it is suitable for spatial data mining.Based on these properties, this thesis discusses the application of several approaches of wavelet analysis in spatial data mining in detail. Firstly, time series data is analyzed by 2D-CWT approach, the analysis of a series of experimentations provid that the approach is effectual. Subsequently, a new wavelet transform based on lifting scheme is present. Due to several properties such as simple computing process, fast calculating speed and lossless transformation, this approach breaks through the previous defect that it improves the efficiency and accuracy...