Research Of Classification Model For Thematic Mapping Data

The broad application of mathematics, in particular modern mathematics, makes great progress in theories and methods of thematic mapping. Statistic thematic mapping is a strong method for analyzing and depicting data, it transforms abstract statistical information to some graphs which can be easily accepted, and the information is endued with geographical meanings. Classification model is the mathematics model frequently used in statistic thematic mapping. Mastering all kinds of classification methods of spatial elements or phenomena, which plays an important effect on building a geographical information system and improving cartography, especially on classifying data, confirming property, editing and managing graph, analyzing data model, and so on.Based on mastering traditional classification model of thematic cartography elements, we apply fuzzying mathematics, statistics, information theory and so forth, to cartographical data processing model, perfect each model of making sure of grading circumscription, and build an algorithm how to automatically fix on the number of grading. The central contents of this article are as follows:First, subsection discretionary numerical sequence clustering method is advanced to replace previous method for its bad precision, after analyzeing each method of data clustering. Because of the shortcoming of complex in superior partition clustering method, a "choosing maximal distance" model is advanced according to the increment of dispersion measure within a cluster.Second, applying fuzzy mathematics to fix on the number of grading, combining with how many clusters you should look for with the data in high dimension, a new fuzzy cover approach based on adaptive inflation factor is found, which is used specially for one dimension data. This approach can not only automatically make sure of the appropriate number of clusters, but also have a comparatively precision clustering result.Third, through making a theoretical analysis in which kind of data every method fits, a new clustering evaluation model is found by judging methods from easy to complex.At last, statistics knowledge is adequately applied to find a new parameter to describe the distribution characteristic which is proved by large numbers of data in VC program. And this parameter can be computed to judge in which kind of grading methods the data fit.