CHOROPLETH MAPS: CLASSIFICATION REVISITED

摘要

Classification was traditionally used as an instrument for producing choropleth maps. In our study we consider the role of the process of classification in exploratory data analysis. We developed a set of tools for classification that facilitate looking on data from various viewpoints and thereby investigate different aspects of the data. The tools include direct manipulation controls for specifying arbitrary class boundaries, graphs representing statistical distribution of attribute values, means for automatic classification, calculation of statistical quality of a classification, and various color schemes that can be applied to represent classes on a choropleth map. The classes can be dynamically propagated to other displays showing different characteristics of the same objects. This helps in exploring relationships between attributes. Additionally, relationships can be investigated by applying various data mining methods to the classes produced. All the instruments are used in a highly interactive and dynamic mode: results of each user's action lead to immediate update of all displays involved.