The classification of GIS objects

Abstract

This paper discusses the current issues of the application of classification and data processing in geoinformation systems. The problems of classification of various objects have been studied in the works of many authors. These include a fairly wide range of problems: decryption of satellite images, pattern recognition, mathematical modeling, etc. In this paper, we study the methods and techniques for classifying objects listed in the literature, as well as preliminary data processing: feature normalization, feature weighting, aggregation, dimensionality reduction, etc. The result of finding spatial features in an attribute space is often a representation of spatial features in the form of an object-feature matrix that reflects the measurement of M features on N spatial features and contains N rows and M columns. To classify spatial objects, you must have a geographical map of these objects and an object-attribute matrix, the rows of which correspond to the spatial objects. In order to properly classify, you need to perform pre-processing of the data, including normalization, weighting, dimensionality reduction, aggregation, and identification. After preliminary data processing, the objects are classified. The paper lists and describes such classification methods as nuclear classification methods, hierarchical divisive classification methods, hierarchical agglomerative classification methods, near neighbor method, far neighbor method, centroid method, group mean method (mean link method) and other issues related to the classification of geoinformation objects.