Task-oriented and change preserving data classification for multi-temporal choropleth maps

Abstract Series or animations of classified choropleth maps are an important form of reproducing multi-temporal, cardinally scaled data sets, especially in media. However, there are problems with such representations that are not taken into account by the known methods of data classification (such as equidistant grouping or quantiles), and therefore lead to visualisations that are not sufficiently suitable for use. On the one hand, different questions and change analyses tasks are not explicitly considered in the process of making these maps. In the following, typical change tasks are singled out (i.e. show absolute differences, absolute percentage changes, positive changes, deviations from the trend) and corresponding metrics for quantitative description are proposed. On the other hand, there are no measures in the usual procedures to avoid the loss of significant changes after classification (i.e. the regions belong to the same class). In the following, a procedure is therefore proposed that begins with rules for assigning value differences to class differences (e.g. based on statistical significance). Based on this, a preservation measure is defined that describes the success of obtaining the desired class differences after applying the classification. This measure can also be used to guide a new classification procedure. Using two multi-temporal data sets, the effects of the developed measures and methods are demonstrated both numerically and visually in corresponding choropleth maps.