Data classification methods for preserving spatial patterns

Abstract. The primary purpose of choropleth maps is to display or even to emphasize special relationships or patterns in the spatial distribution of attribute values. However, because classification methods commonly used and implemented in software packages (such as equidistance, quantiles, Jenks, etc.) are data-driven, a preservation of such spatial patterns is not guaranteed. Instead of such a data-driven approach in the following a task-oriented procedure is pursued: For typical patterns (local and global extreme values, large value differences to neighbours, spatial clusters, hot/cold spots) specific algorithms have been developed, implemented and tested.

Download Full-text

A Data-Driven Approach to Determine the Single Droplet Post-Impingement Pattern on a Dry Wall Using Statistical Machine Learning Classification Methods

10.4271/2021-01-0552 ◽

2021 ◽

Author(s):

Jiachen Zhai ◽

Seong-Young Lee

Keyword(s):

Machine Learning ◽

Data Driven ◽

Classification Methods ◽

Single Droplet ◽

Statistical Machine Learning ◽

Machine Learning Classification ◽

Data Driven Approach

Download Full-text

Alexa Conversations: An Extensible Data-driven Approach for Building Task-oriented Dialogue Systems

10.18653/v1/2021.naacl-demos.15 ◽

2021 ◽

Author(s):

Anish Acharya ◽

Suranjit Adhikari ◽

Sanchit Agarwal ◽

Vincent Auvray ◽

Nehal Belgamwar ◽

...

Keyword(s):

Data Driven ◽

Dialogue Systems ◽

Data Driven Approach ◽

Task Oriented

Download Full-text

Task-Oriented Visualization Approaches for Landscape and Urban Change Analysis

ISPRS International Journal of Geo-Information ◽

10.3390/ijgi7080288 ◽

2018 ◽

Vol 7 (8) ◽

pp. 288 ◽

Cited By ~ 2

Author(s):

Jochen Schiewe

Keyword(s):

Extreme Values ◽

Urban Change ◽

Future Research ◽

Complex Data ◽

Visual Methods ◽

Change Analysis ◽

Value Differences ◽

Geovisual Analytics ◽

Manual Processing ◽

Task Oriented

Approaches to landscape and urban change analysis are still far away from being fully automatic or operational. For this reason, the concept of Geovisual Analytics is proposed, combining computational and visual/manual processing steps. This contribution concentrates on the latter part with the overall goal of improving its usability. For this purpose, a classification of tasks is created, which often occur in the context of change analysis. This serves as the basis for the assignment of suitable map types to change processing results. Beyond this, it is pointed out that in many cases an appropriate pre-processing of data is imperative to preserve or enhance certain spatial relationships or characteristics for visualization. This is demonstrated using the example of data classification prior to choropleth mapping. Methods are described which allow the preservation of local extreme values, large value differences between adjacent polygons, clusters, and hot/cold spots. Finally, discussing future research and developments, it will be stressed that the importance of visual methods in the context of big data change analysis will continue to increase, which is due to the particular ability of maps to generalize and reduce complex data to a minimum.

Download Full-text